America's Habit

Drug Abuse, Drug Trafficking, & Organized Crime

President's Commission on Organized Crime, 1986

Consultant Paper: Data and Analysis Requirements for Policy Toward Drug Enforcement and Organized Crime

By Mark Kleiman. Submitted: November 1985. Mark Kleiman is a Research Fellow in the Program in Criminal Justice Policy and Management at the John F. Kennedy School of Government of Harvard University, Senior Lecturer in Political Science at the University of Rochester, and President of BOTEC Analysis Corporation.

Executive Summary

Organized crime control and drug enforcement need to be considered together insofar as drug trafficking contributes to the organized crime problem or organized crime contributes to the drug problem. In making policy it would be useful to know about (1) the role of drug trafficking in supporting organized crime groups; (2) the role of organized crime groups in supplying drugs; and (3) the effects on each problem of changing enforcement practices in the other.

Drug trafficking is almost certainly the largest source of illegally-earned income in the United States.

Both "traditionally" organized crime groups (Mafia families) and "emerging" organized crime groups participate in some aspects of the illicit drug trafficking industry. In addition, some drug trafficking organizations have some of the characteristics of organized crime groups, such as the capacity and reputation for the use of violence and corruption to frustrate enforcement.

Organized crime can participate in drug markets either "productively," using its lawbreaking skills to contribute to the supply of illicit drugs, or "parasitically," using violence and the threat of violence to extort drug merchants or organize cartels. Only the productive activity of organized crime contributes to the drug supply problem, which is the main target of drug enforcement. Parasitic organized crime activity, in illicit markets as in legitimate markets, tends to raise prices and reduce consumption.

Controlling organized crime and controlling drug abuse (and drug-abuse-related crime) are two distinct objectives. Some agencies of government - the Drug Enforcement Administration (DEA) and the National Institute on Drug Abuse (NIDA), for example - have primary responsibility for drug abuse control. Others, like the Organized Crime Strike Forces, are concerned primarily with organized crime control. The two sets of agencies may or may not promote each other's objectives.

Drug enforcement actions may affect the drug markets in several ways. They may: (1) help or hurt the efforts of organized crime to participate in and profit from drug dealing; (2) encourage or retard the development of organized-crime-like characteristics among drug trafficking groups; and (3) lead to more or fewer convictions of organized criminals. Unless it is deliberately focused on organized crime groups, high-level drug enforcement may tend to foster organized crime by keeping prices high and driving out less well-organized competitors.

Organized crime enforcement actions may make the task of drug enforcement harder or easier, depending on whether the roles of those groups in the drug markets are primarily parasitic or primarily productive. Aside from heroin importation into New York, traditional organized crime is largely a parasite on the drug markets; thus, as long as the organized crime effort concentrates on the Mafia, it has little to contribute to drug supply reduction.

Data about drug trafficking and drug enforcement, and about the relationships between the drug trade and organized crime, can serve three purposes: describing the current dimensions of the problem, reasoning about how changes in enforcement activity might make matters better or worse, and measuring the actual effects of such changes after they are made.

From the perspective of organized crime policy, essential facts about the drug markets are prices and quantities, and, most important, the relationship between changes in retail price and changes in the quantity purchased: the price-elasticity of demand. Driving up the price of an inelastically-demanded drug will increase the total revenues of traffickers of that drug, and thus tend to promote organized crime.

Data about the drug markets and drug enforcement come from three sources: (1) survey research; (2) observations of the consequences of drug abuse (deaths, injuries, and demand for treatment); and (3) enforcement statistics and reports.

Current large survey efforts, among households and high-school seniors, do not gather much price and quantity data. We need a separate national survey with a sample designed to overrepresent heavy drug-users, that ask very detailed questions about price, quantity, and conditions of drug purchase. In addition, local and regional surveys should be conducted to evaluate the effect of enforcement initiatives on market conditions and drug consumption.

Two techniques not now in use - "panel" surveys of drug users and surveys of convicted dealers - should be undertaken, if the difficult operational and ethical issues can be resolved.

Neither abuse-consequence data now gathered (medical examiner and emergency-room reports through the Drug Abuse Warning Network (DAWN), drug abuse treatment data through the remnants of the Client-Oriented Data Acquisition Process (CODAP), opinions of treatment professionals through the Community Epidemiology Work Group (CEWG)) nor alternative ways of reaching the same facts are very useful in making organized crime/drug enforcement policy choices. These programs serve other purposes and should not be disturbed, except for adding bioassay capability to some DAWN sites. Pilot programs to test jail admittees for heroin use are worth pursuing.

More could be done with enforcement agency data, which could tell us about the enforcement effort itself (agency activities, targets, measured results) and on its effects (prices at different levels of the drug trade). The new management of the Drug Enforcement Administration has put less effort into estimating prices and quantities in the drug markets. That change should be reversed, and DEA should again publish price-chain estimates.

Enforcement agencies can also make and analyze retail-level drug purchases, unrelated to specific cases, to measure the state of the drug markets. Such "monitor" programs should be increased in frequency, systematized, and applied to marijuana and cocaine as well as heroin.

Recent revisions in the methods used by the National Narcotics Intelligence Consumers Committee (NNICC) in preparing the annual Narcotics Intelligence Estimate (NIE) appear to have somewhat reduced that document's inconsistencies with observable fact. However, the NNICC still seems to practice analysis by negotiation rather than basing its estimates on evidence. The NIE would be more useful if its assumptions, data, and computations were explicit. There might be value in allowing a group of experts outside the government access to NNICC documents and sources, from which they could develop an independent set of estimates.

Questions to be Answered:

Drug trafficking and organized crime are linked problems. Policy making and evaluation for each is complicated by the presence of the other. My objective is to outline a set of data collection and analysis activities that would answer questions under four headings:

1. What effects do organized crime have on drug trafficking and enforcement?

2. What effects do drug trafficking and enforcement have on organized crime?

3. How should federal agencies engaged in organized crime enforcement adjust their policies to contribute to the objectives of drug policy?

4. How should federal agencies engaged in drug enforcement adjust their policies to contribute to the objective of organized crime control?

Data collection and analysis are never free; in the drug area, they are likely to be quite expensive. It is therefore important to identify in advance what is relatively certain on the basis of information and analysis now in hand; it doesn't take a spectroscopic analysis to discover that the sky is blue. The following section will therefore attempt to describe what we now know and to identify the key analytic points.

Introduction: Drug Enforcement and Organized Crime Control

How Two Objectives and Two Sets of Institutions Intersect

The objective of drug enforcement is to keep drugs away from consumers. The objective of organized crime enforcement is to control the wealth and power of major criminal organizations and to frustrate their goal of being able to defy the law.

These objectives interact where the illicit drug business is a major source of power or revenue to organized crime groups (or a major source of their vulnerability to enforcement), and where the capabilities of organized crime groups make drugs more available than they would otherwise be: in other words, where the drug trade contributes to the organized crime problem or where organized crime contributes to the drug problem.

Defining Organized Crime

The term organized crime can be used in one of two ways. It may be merely an ethnically neutral euphemism for "the Mafia" or "La Cosa Nostra" (LCN); or it may describe a class of criminal organizations whose characteristics make them particularly worthy of law enforcement attention. Size, wealth, political power, participation in a variety of criminal activities, and continuity over time have all been suggested as such characteristics.

One way to bridge the gap between the two definitional approaches is to refer to the Mafia as "traditional" organized crime, while acknowledging the existence of other, "emerging," organized crime groups.

Defining organized crime in terms of its characteristics, rather than ostensibly, complicates the analysis. Not only does this approach treat the set of "organized crime" groups as variable rather than fixed; it also implies a many-valued rather than two-valued classification. Criminal groups, rather than simply being organized crime or non-organized crime (i.e., Mafia families or other), will fall along a spectrum of being more or less "organized-crime-like." This suggests the possibility that drug enforcement activity can increase or decrease the number of drug-dealing organizations with many organized crime characteristics. It also suggests that it can make existing organized-crime-like drug dealers more or less organized-crime-like and give such groups a greater or lesser share of the drug market.

This paper will attempt to distinguish between traditional and emerging organized crime groups, but will think of organized crime as primarily a set of organizational characteristics rather than a predefined set of organizations. The relevant characteristics are organizational continuity and organizational investment in the capability and reputation for violent or corrupt interference with law enforcement activity.

Organized Crime as a Drug Supply Problem

"Traditional" organized crime - La Cosa Nostra - does not now contribute substantially to the drug problem, with the exception of the major roles of two of the five New York LCN "families" in heroin importation. This is not to say that LCN members and associates do not participate in drug dealing on many levels, but only that they make no distinctive contribution to the supply of illicit drugs. Indeed, as Thomas Schelling has pointed out, the activity of organized crime in cartel formation and extortion from illicit entrepreneurs may on balance reduce the supply and increase the price of illicit goods, as is true in licit markets.

As a consequence, organized crime enforcement institutions - the Strike Forces and the investigators who work with them - have relatively little to contribute to solving the problem of illicit drug supply, as long as they continue their concentration on "traditional" organized crime groups. This concentration has been so successful, both in terms of successful cases and in terms of creating and maintaining productive, high-morale enforcement and prosecution institutions, that it seems both unlikely (as a matter of prediction) to be changed and unwise (as a matter of policy analysis) to consider changing.

It may be the case that some "emerging" organized crime groups make distinctive contributions to various parts of drug supply, so that their disruption would contribute to drug supply control (over and above the general effect of enforcement in raising prices by raising the costs and risks of drug trafficking). Outlaw motorcycle gangs - the Hell's Angels in the West, the Outlaws in the Mid-Atlantic region - may be "least-cost suppliers" in the distribution of cocaine, amphetamines, and phencyclidine (PCP). Similarly, some ethnic-based organizations among Asian-Americans may be significant contributors to the heroin supply, as the Herrera organization certainly was at least until the recent arrests. By contrast, neither the importation nor the distribution of marijuana appears to offer the economies of organization that would make breaking up particular trafficking groups a major marijuana enforcement objective. The marijuana trade may be at its most efficient as a decentralized set of independent buyers and sellers.

To develop a set of institutions to appropriately address the role of emerging organized crime groups in the drug markets will require devising organizational incentives and career patterns that support extremely long-term investigations and investments in intelligence (broadly defined), while maintaining a supply-control orientation among agents and prosecutors. This will not be an easy task. There may be lessons to be learned from the Federal Bureau of Narcotics in its approach to LCN heroin dealing.

The Drug Markets as an Organized Crime Problem

The illicit drug traffic contributes to the organized crime problem because of the enormous amounts of money at stake. These potential revenues will both attract (and help to sustain) existing organized crime groups and call into existence new groups with organized crime characteristics.

LCN families, other ethnically-based organizations long in existence, and outlaw motorcycle gangs may engage in drug dealing as a way to support themselves financially and to take advantage of their organizational abilities in illicit transactions. The opportunities in the drug trade may also allow the development of organizations whose origin is in drug dealing, and which may remain specialized in drug dealing, while taking on organized crime characteristics, in particular, organizational continuity and organizational investments in the abilities to use violence and corruption to resist enforcement efforts while remaining prominent in illicit markets.

Drug trafficking is almost certainly the largest source of illegally-earned income in the United States, far outstripping the market for illegal gambling or prostitution. Estimates of total revenues from drug sales vary widely and are highly unreliable. It seems unlikely that total annual revenues are much less than $25-30 billion.

Virtually all of this represents rewards for lawbreaking; the payments forlicit goods and services (rental or purchase of vehicles and buildings; lawyers' fees) are negligible. On the other hand, relatively little represents revenue to large criminal organizations, because so much of the final retail price of illegal drugs consists of the markups of retail dealers and low-level wholesalers. For heroin and cocaine less than one-tenth of the final retail price is collected by the dealer who sells kilogram units. Marijuana, by contrast, with much less vigorous enforcement at lower levels, undergoes only about a four-to-one price markup between first domestic distribution and final retail sale.

Therefore, while all three markets have total values on the order of $10 billion per year, more than $2 billion of the marijuana market represents the revenues of high-level organizations, while the heroin and cocaine markets at that level represent about $1 billion each. On the other hand, the tighter distribution channels for heroin and cocaine make it possible for large and highly disciplined organizations to exist at the wholesale-retail level, so that organizations with many of the characteristics of organized crime may exist at lower levels of those trades (e.g., the Barnes organization in the Harlem heroin market).

Making Strategy

Drug enforcement actions change the drug markets in ways that may: (1) help or hurt the efforts of existing organized crime enterprises, whether traditional or emerging, to participate in and profit from them; (2) encourage or retard the development of organized crime characteristics in existing or new enterprises within the drug markets; (3) be more or less helpful in securing convictions against members of already-identified organized crime groups and the identification of new groups meeting those criteria.

The more effectively drug enforcement is directed toward organized crime control, the less hospitable organized crime groups will find the drug markets. On the other hand, an increase in the overall level of drug enforcement not aimed specifically at organized crime groups may tend to frustrate organized crime control by helping existing organized crime groups profit from the drug trade and by encouraging existing drug enterprises to take on more organized crime characteristics.

The ability of organized crime groups to profit from the drug trade - and the tendency of the drug trade to spawn new organized crime groups - depends both on the amount of money to be made in that trade and on whether the environment rewards or punishes tight control and a reputation for violence. Under some conditions, tightened enforcement can both boost the total money available to be made and confer competitive advantage on more violent groups at the expense of less violent ones.

Tighter enforcement can increase the money to be made in any black market, if it succeeds in raising the price of the commodity involved and if consumption decreases less, in percentage terms, than prices rise. (A 50 percent price increase that causes only a 20 percent decrease in volume will create a new market with total revenues 1.5 x 0.8 = 1.2 times as large as the revenues in the old market, and thus increase the money to be made by 0.2 or 20 percent.) Thus the sensitivity of consumer demand to retail price - what economists call the price-elasticity of demand - becomes an essential number to know.

If the price-elasticity of demand is (in absolute value) greater than one - if a given percentage increase in price will cause a larger percentage decrease in volume - then price increases will decrease total revenues. (Such goods are said to be "relatively elastically demanded.") On the other hand, if the price-elasticity of demand is (in absolute value) less than one - if a given percentage change in price will cause a smaller percentage change in volume, as in the example above - then price increases will increase total revenues. (Such goods are said to be "relatively inelastically demanded.")

In markets where demand is relatively elastic - where consumers will respond to price increases by cutting back consumption sharply - increased enforcement will reduce the money at stake and thus tend to make the markets less lucrative for organized crime. Where demand is inelastic, increasing enforcement will increase the money at stake and thus make the markets more able to support organized crime. Much of the empirical work recommended below centers on measuring the price-elasticity of demand for various drugs.

Unfortunately, it appears that the demand for marijuana and cocaine is relatively inelastic. (The argument with respect to marijuana is on much firmer ground than the argument with respect to cocaine.) In any case, tightened general drug enforcement - as opposed to drug enforcement targeted at organized crime - will tend to give organized crime groups and organized crime-like groups relative competitive advantage, because their capacity for violence and corruption helps protect then against ordinary enforcement actions. This is the central tension in organized crime/drug enforcement policy: good drug enforcement may be bad organized crime control.

Observations to be Made

The observations to be made fall into three broad classes: retail purchasing and consumption, lower-level dealing, and enforcement activity. (Of course, observations on high-level dealing are also of great importance, but they are almost impossible to distinguish from enforcement intelligence. Data collection here means talking to agents, enforcement-agency analysts, and prosecutors, and the information to be collected is rarely in quantitative form.)

Three Objectives of Data Collection and Analysis

Data collection and analysis about drug markets can serve three distinct functions:

- Descriptive: measuring the current dimensions of the problem and identifying trends. Sample questions: How much cocaine was sold to end users last year? What prices do they pay? How were those revenues distributed among individuals and groups in the illicit cocaine industry? What were the relationships between identified organized crime groups and cocaine trafficking? How have the volumes and prices in the cocaine market changed over time? What organized crime groups are active in the cocaine market? What role does it play in their other activities? What roles do they play - productive or parasitic - in the market?

- Theoretical: generating and testing hypotheses about relationships among variables, and in particular about the likely effects of changes in enforcement. Sample questions: Are there "barriers to entry" at some levels of the cocaine trade that allow traffickers to collect "pure" profits ("economic rents")? How would the average size and longevity of cocaine-trafficking organizations change if enforcement were substantially tightened?

- Evaluative: measuring the actual effects of enforcement initiatives. Sample question: How did the South Florida Task Force change prices and volumes in the cocaine market?

Three Sources of Data

Illicit activities are by their nature hard to observe and measure, because participants attempt to conceal them from authorities. Excise tax data, which provide the most convenient measures of alcohol and tobacco consumption, are unavailable for illicit drug markets. The same is true of the reports routinely prepared by licit enterprises reports to state and federal tax collectors, to trade associations and government statistical collectors, to trade associations and government statistical agencies, and to stockholders - from which estimates of licit markets are derived.

The remaining potential sources of information about illicit drug markets are:

(1) Surveys of drug users and potential drug users;

(2) Observations of the consequences of drug - abuse health effects, intoxicated behavior, and demand for treatment - and the opinions of professionals who deal with those consequences; and

(3) Enforcement data and the opinions of agents and prosecutors.

Note that while surveys and observations of consequences tell us almost exclusively about the behavior of drug consumers and drug dealers, enforcement data and the opinions of enforcement professionals can also tell us about the other half of the equation: the governmental response.

Several current programs now attempt to tap these three sources of information. None is without major flaws from the perspective of organized crime policy. A major problem common to many is that they are more concerned with numbers of users and frequency of use consumed than with quantities consumed and flows of revenues.

Survey Research

Survey research is potentially the most powerful tool available for studying the markets for illicit drugs. Unlike data from enforcement activity or abuse consequences, survey research can attempt to measure prices and quantities directly rather than through complicated chains of inference. Performing surveys about consumption activities - market research supports a rather prosperous industry, including both for-profit enterprises and prestigious universities and nonprofit research centers. These institutions have developed a repertoire of established techniques. The fact that consumer-goods firms continue to use it heavily testifies to the ability of market research to measure and predict consumer behavior.

The National Institute on Drug Abuse funds two series of survey research projects on illicit drug consumption. The Social Research Group of George Washington University does a survey every two years of more than 5,000 persons 12 years or older living in households [Miller et al. 1983]. This series is referred to among students of drug abuse as "the household survey."

Every year, the Institute of Social Research of the University of Michigan questions about 17,000 seniors in high school. The Michigan reports are referred to as "the high-school senior survey." These two surveys are believed to cost NIDA something less than one-half a million dollars for each two-year cycle.

The concept of survey research is simple. Draw a sample so designed that every member of the population under study has an equal probability of being chosen. (This is what statisticians mean by the term "random sample." A variant on simple randomization is to divide the population under study into subgroups, called "strata," and to draw the sample so that each stratum is represented in proportion to its representation in the population, and each member of a given stratum has the same chance of being drawn as any other member. This procedure, called "stratified random sampling," is the most commonly used.) Having drawn the sample, ask questions and record the answers. Then use the percentage answers from the sample as estimates of the behavior of the population from which the sample was drawn: that is, make population projections based on sample results.

Surveys can be inaccurate for three reasons: random error, sample bias, and response bias.

Random error (or "sampling error") occurs because the law of averages only works on average. If 30 percent of the population uses Crest Toothpaste, then in a sample of 100 persons drawn at random about 30 will say they use Crest. But the operative word is "about:" one time it will be 28, another 33. The larger the sample, the smaller the random sampling error, in accord with well-known statistical formulas. This does not depend on the size of the population to be measured: a sample of 1,000 is subject to the same sampling error in estimating a population of 100,000 as in estimating a population of 10 million. The rule of thumb is that the sampling error is inversely proportional to the square root of the sample size: a sample four times as big gives estimates one half as uncertain. For many purposes - predicting the outcome of a Presidential election, for example - a sample size of a few hundred is ample.

The smaller the percentage of a population that has some trait, the larger the sample needed to measure that trait accurately, because a sampling error of 2 percent looms larger if the underlying figure is 5 percent than if it is 50 percent. Thus, attempts to measure rare phenomena - heroin use, for example - take bigger samples than attempts to measure common phenomena such as alcohol use. The other factor that drives up needed sample sizes is the desire to study how subgroups of a population differ. In a study of the effect of income on cocaine use, the relevant sample size is the number of respondents within each income class. To study whether recent increases in marijuana enforcement by the Boston DEA office succeeded in reducing marijuana consumption, one would need a sample including at least a few hundred marijuana consumers from New England, and a similar sample from another area to serve as a control.

Other than the expensive expedient of using very large sample sizes - such as the 50,000 households interviewed each month by the Bureau of Labor Statistics - the only way to reduce sampling-error problems in estimating rare traits is to deviate from stratified random sampling. Areas under special study or subpopulations among which the trait in question is less rare can be deliberately overrepresented. For example, if cocaine use is concentrated among upper-income urban males in the 20-45 age bracket, a study of cocaine use might deliberately seek out a sample including more such people; this is called oversampling. The results of a survey using oversampling need to be adjusted to project them back on the population. Those adjustments introduce great complexity into the statistical analysis of survey data. Subtle errors can easily creep in. In consequence, survey researchers try to avoid using oversamples.

Random error, just because it is random, does not tend to bias results. The luck of the draw produces overestimates and underestimates impartially. It does, however, introduce uncertainty. This is more of a problem when trying to measure trends (where small changes from one survey to the next matter greatly) than in measuring absolute levels. By contrast, both sample bias and response bias can create systematic divergences between survey results and actual phenomena.

Sample bias comes from a failure to draw a representative sample. If the sampling technique makes persons with a given trait under study more or less likely than average to be selected, the sample results will consistently overestimate or underestimate the trait in question. For example, the household survey omits persons with no fixed address and those living in dormitories, barracks, and prisons. If very heavy drug users are more likely than the general population to be drifting, away at school, in the service, or in prison, then the household survey will underestimate the number of very heavy drug users. Similarly, the high-school senior survey may underrepresent drug abuse among persons in that age range for two reasons. By choosing to study high-school seniors, it omits dropouts, who may be more likely than average to be heavy drug users. In addition, the actual sampling technique may tend to underrepresent frequent absentees, and frequent absenteeism seems to be associated with heavy drug use.

If the only statistics of interest are population proportions (e.g., the proportion of the voters who think the President is doing a good job), sample bias is a controllable problem. When frequency or quantity is at issue (e.g., how frequently the average high-school senior smokes marijuana, how much cocaine was consumed last year, the total amount paid for heroin) sample bias can be a major worry. A small proportion of any population can easily account for a very large fraction of any activity, from baseball games attended to books purchased. In the case of alcohol, for example, 5 percent of the population accounts for 50 percent of the consumption. By one estimate, 10 percent of the marijuana-using population smokes two-thirds of the marijuana. Cocaine consumption may be even more skewed, with the additional problem that the total number of cocaine users is itself much smaller than the totals for marijuana and alcohol; two-thirds or more of the cocaine consumption may be accounted for by one-half a million users.

Sampling procedures to guard against underrepresenting such very small populations are hard to devise, particularly when the behavior under study tends to make them furtive or to reduce their participation in various normal activities. Oversampling can help; so can what are called "snowball" techniques, where respondents who fit some profile are asked for the names of other respondents. The illicit nature of drug abuse can make "snowball" sampling very difficult, however.

Response bias comes from the systematic failure of respondents to answer accurately. Any survey will gather some inaccurate responses, due to misunderstanding the questions, lapses of memory, or deliberate deception by respondents resulting from fear of breach of confidentiality, embarrassment, or a simple desire to tell the interviewer what the respondent thinks he or she wants to be told. As long as those errors cancel out, they pose no problem. The problem comes if the inaccuracies are tilted to one side or the other. A survey that asks questions about activities that are both illegal and socially disapproved runs a greater-than-normal risk of encountering response bias.

A clue to how serious response bias can be, even with legal behavior, comes from surveys where people are asked how much alcohol they drink. Only two-thirds of the alcohol on which excise tax is paid (which may be an undercount of alcohol sold if distillers cheat the Treasury) is accounted for by survey responses; the survey data thus underestimate consumption by at least 50 percent. It has been suggested that psychological "denial" by a small population of extremely heavy drinkers may account for most of the difference. Here again, as with sample bias, problems with a very small portion of the sample can cause the results to be seriously wrong. On the other hand, a compilation of studies in which survey responses to questions about illegal drug use were compared with urine-test results showed no tendency to underreport.

A special kind of response-bias problem comes when survey data are used to estimate trends. If the level or social disapproval as perceived by recipients changes, then response bias may change too. For example, increasingly negative attitudes about marijuana use might lead adolescent respondents to the high-school seniors survey to be less likely to tell adult interviewers that they smoked marijuana every day, even if their actual behavior did not change. The result would be an apparent, but spurious, decline in heavy marijuana use among adolescents. (This is not to say that changing attitudes might not also change behavior, only that the survey evidence is subject to error from this source).

There are no easy answers to the problem of response bias. Sometimes, as in the studies cited by K. H. Marquis, it is possible to collect independent data as a check on a sample of responses, and then adjust the sample results for the measured bias. Another strategy is to vary the circumstances of response, for example, by using "secret ballot" answer sheets instead of face-to-face questioning or by changing the characteristics of the interviewers. Just as pollsters long ago discovered that black respondents answered many questions differently when the interviewers were black, it may be the case that some adolescents would respond differently to questions about drug use put by another adolescent - particularly one dressed and groomed in ways consistent with drug use - than to the same questions put by an adult.

Changes and Supplements to Current Surveys

Both the household survey and the high-school seniors survey, conducted by experienced researchers at institutions with reputations for producing high-quality survey research, are well-designed for their intended purposes: primarily the measurement of the number of drug users, secondarily the measurement of frequency of drug use. But from the perspective of making drug enforcement policy, and particularly policy to control organized crime, they do not gather the right data, or gather it from the right people, or gather it frequently enough. The most important things surveys can tell us are the quantity of each drug purchased and consumed; the prices at which those purchases are made; the identities and relationships of buyers and sellers; the effects of enforcement on market conditions; and the effects of market conditions on consumption. In order to learn these things, surveys would have to:

1. Find More of the Heavy Drug-Users. So much of total drug consumption is concentrated among a relatively few very heavy users that no practical increase in sample size for a stratified random sample among the general population will yield an adequate sample. For example, the 5,000-person household survey could expect to get by chance about a dozen of the one half-million heaviest cocaine users. There are two ways to get a meaningful sample of heavy users: either use oversampling in drawing respondents for the household and high-school senior surveys, or run separate studies of heavy use. The "Young Men and Drugs" studies show both the feasibility of finding samples rich in serious drug abusers and the value of the results that finding such samples can yield.

2. Ask about Prices and Quantities purchased. We would like to know when the user's last purchase of a given drug took place; what quantity was involved, plus whatever the purchaser knows or was told about purity/potency/ quality; and how much money changed hands; this information is essential to making any serious estimate of the dollar volume of the market.

3. Ask about the Conditions of the Transaction. Where did the purchase take place? (The purchaser's residence, the seller's, a school building, an office building, an apartment corridor, a park, a street corner?) Were there other people present? Had the transaction been prenegotiated? Was the price take-it-or-leave-it, or was there negotiation? Was the purchaser's previous transaction with the same seller or a different seller? Is the seller a relative, a close friend, and acquaintance, or a stranger? These questions can help answer questions about the possible effectiveness of retail-level enforcement as a way to cut down on the volume of drugs consumed.

4. Ask about Price Increases, Shortages, Sharing, and Consumer Inventories. Was the price of the most recent transaction higher or lower than the respondent's previous transaction? Than the respondent expected? If the price of the drug were to increase by 25 percent, what would the respondent do? (Use about the same amount, use less, quit altogether, switch to another drug?) Has the respondent ever been unable to find a supply of the drug in question at any price? Does the respondent now have a supply of the drug? If so, how long would it last at his normal consumption rate? How frequently has the respondent used drugs purchased by another? Given drugs to another? These questions will help determine the effect of price increases and temporary shortages on consumption behavior. Unfortunately, experience from market research on legitimate products suggests that the answers to the what-if questions are likely to be unreliable.

5. Give Information about Local Markets. Federal and local enforcement efforts vary widely across the country in their intensity and success, but there are almost no studies of what the effects are on retail purchases and consumption. Nationwide samples need to be supplemented with local ones.

6. Provide a Steady Stream of Data in Addition to Annual or Once-Every-Two-Years Snapshots. This is particularly valuable in combination with local data as a way of measuring the difference enforcement makes (or fails to make). Is the availability of drugs fairly constant, or does it vary? How do consumers respond? Estimating the price-elasticity of demand requires frequent measurements of both price and quantity, both to generate an adequate number of data points and to avoid confounding price effects with long-term trends or cohort effects on the demand side.

One alternative to random sampling often used in market research is a panel survey, where questions are asked repeatedly of the same group of respondents. Panel data would be enormously useful in understanding the effects of market conditions on drug use, and the individual patterns of drug consumption. For example, when drug supplies are short, how are consumption decreases distributed among new users and existing users, poor users and affluent users, heavy users and light users? What drugs are used as substitutes when marijuana is scarce or expensive? How many users, who cease use when supplies are tight, never start up again, or start up at lower levels of use? An obvious problem is assuring respondent confidentiality, and convincing respondents that it is assured. Perhaps more worrisome, being members of such a panel might make respondents more drug-conscious, which raises serious ethical questions, as well as threatening the validity of such studies.

Obstacles

Naturally, proposing such studies for someone else to fund and carry out is much easier than doing the work. Aside from the problems of expense and the mass of legal, ethical, and operational issues raised by asking detailed questions about illegal transactions, such studies would face several obstacles:

First, they would have to be cleared by the Office of Management and Budget, which views any expansion of survey research by the government with a jaundiced eye. The complexity of the questions to be asked strongly argues for the value of paying respondents a small fee to encourage them to complete the interview, but this makes OMB reviewers even more nervous. (The 1981 household survey didn't get into the field until 1982 because of a dispute with OMB about forms and payments to respondents.)

Second, modifying the existing surveys in terms of content or sample selection will interfere with the ability of researchers to compare results from one survey to the next to determine trends. The preference of survey researchers for asking the same questions in the same words to the same samples is not mindless conservatism, but a well-founded belief in the virtues of repeating measurements exactly, rather than changing survey series represents a big investment of money and talent.

Finally, it is worth noting that neither the National Institute on Drug Abuse, which funds the current surveys, nor the researches who conduct them, share the policy and research perspective urged here. NIDA is far more interested in the patterns of drug abuse than in the size and functioning of the drug markets. Nor is there any demand for this sort of research from the enforcement and prosecution agencies.

Surveys of Dealers

The discussion above concerns only surveys of drug users. It is also possible to survey drug dealers, asking both about their own dealings - transactions, strategies for dealing with customers, suppliers, and enforcement agencies, incomes, expenses, encounters, if any, with organized crime, opinions of enforcement risk, pricing policy, recruitment into the trade, career patterns and goals - and about their knowledge of how the drug markets work. There is probably no better way to start on serious estimates of, for example, bow many marijuana or cocaine to deliver drugs to the market - is more plausible the fewer the participants, and the more stable the organizational structure, of any given market.

Finding and interviewing a sample of dealers poses both problems and risks for the researcher. However, we do have access to a sample of drug dealers with nothing better to do than answer questions: they are in federal and state prisons. Of course, they may be unrepresentative of the overall dealer population, particularly in terms of their success in avoiding enforcement, and may also be unwilling to be interviewed, or to speak candidly.

Still, if systematic interviews could be conducted with a sample of convicted drug dealers, the results could not fail to be informative. The Rand Corporation is now conducting a pilot study along those lines funded by the National Institute of Justice.

Consequence Data

One alternative to asking people about their consumption of illicit substances is to observe the consequences of drug abuse. These consequences include coming into contact with both the police (arrests, automobile accidents) and the health care system: hospital emergency rooms, medical examiners' offices, and drug treatment facilities. There is now no systematic attempt to record drug intoxication at arrest or at accident scenes of drug users, but the National Institute of Drug Abuse maintains three collection activities to gather medical data: the Drug Abuse Warning Network (DAWN), the Client-Oriented Data Acquisition Process (CODAP) and the Community Epidemiology Work Group (CEWG).

The Drug Abuse Warning Network (DAWN)

The National Institute on Drug Abuse collects and publishes [National Institute on Drug Abuse Statistical Series I) information from about one-sixth of the nation's emergency rooms, with a heavy concentration on 26 metropolitan areas (SMSA's), and from about two-thirds of the medical examiners in those 26 SMSA's only. The emergency rooms report on patients who come in as a result of acute ill experiences with the non-medical use of drugs; the medical examiners report on persons who, in their judgment, died as the direct result of drug use. In 1981, the most recent year for which the data have been published, the DAWN system recorded 121,268 emergency room visits and 2,825 deaths, as a result of drugs. (The deaths exclude New York City, which did not report to DAWN in 1981; in previous years, New York had accounted for about one-quarter of DAWN medical examiner reports.)

DAWN records all drugs used incidents of multiple drug abuse: thus 121,268 emergency room "episodes" yielded 199,400 "mentions," and 2,825 deaths involved 5,640 mentions. Heroin is the third most frequently-mentioned drug in emergency rooms, after alcohol-in-combination (alcohol alone is now counted in DAWN) and diazepam (Valium). Cocaine is sixth, marijuana seventh. Others in the top ten are aspirin, methaqualone, acetaminophen (Tylenol), flurezapam (another tranquilizer) and PCP. Heroin, marijuana, and cocaine, the three largest drug markets in dollar terms, account for fewer than 20,000 total mentions among them. This is such a tiny proportion of the users of these drugs (about 1 in 50 heroin users, about 1 in 1,000 cocaine users, about 1 in 5,000 marijuana users) as to make any inferences from the DAWN sample to the general population extremely questionable. In any case, other than personal demographic data, no questions are asked of DAWN subjects other than "what did you take?" For some drugs, even the answer to this question may be open to doubt, since illicit drug sellers frequently misrepresent their wares, and DAWN sites are not equipped to assay blood or urine to determine the chemical composition of the "speed" or "ludes" that subjects report they have taken.

Attempts to base estimates of patterns and trends in drug availability on DAWN data, but they have not been convincing. DAWN is valuable as an early warning network about both street drugs and pharmaceuticals causing injury and death; but it does not, and almost certainly cannot, contribute much to our understanding of the kind of drug trafficking that contributes to organized crime.

The Client Oriented Data Acquisition Process (CODAP)

Through 1981 NIDA gathered data from all federally-funded drug treatment clinics and programs and published them as NIDA Statistical Series E. Incentive to participate was provided by making participation a condition of funding under the NIDA categorical grant program that paid for drug treatment. The abolition of that categorical grant program and the inclusion of drug treatment in the Community Development Block Grant program of direct fiscal assistance to the states both reduced the total number of drug treatment facilities and eliminated any direct relationship between the clinics and NIDA. As a result, NIDA can mo longer compel participation in CODAP. Approximately 15 states have voluntarily continued to submit CODAP data, and NIDA collects the information and prepares am unofficial series of annual reports. In 1981 there were 249,206 admissions to CODAP facilities, more than three-fourths for outpatient treatment. (Persons admitted more than once during the year were double-counted, so the number of individuals admitted was somewhat smaller). At any one time there were an average of 126,000 persons under treatment, almost 85 percent as outpatients.

While CODAP collected demographic data and information about the drugs for which its clients sought or were assigned to treatment, it did not collect data about individual purchase and consumption patterns. Variations in CODAP counts over time might indicate either increases in abuser populations brought about by increased availability, growing numbers of addicts without connections brought about by decreased availability, or merely increased availability of treatment. CODAP data were of great interest to treatment professionals and produced interesting time-series data on such indicators as age at first use (which allowed some inferences about "cohort effects"). However, CODAP had even less direct relevance to enforcement planning than did DAWN. While local studies of enforcement effectiveness may wish to gather data on treatment demand, this can probably be done as effectively on an ad hoc basis as through a process like CODAP. Reviving that process would not seem to be a high priority from the viewpoint of organized-crime policymaking.

The Community Epidemiology Work Group (CEWG)

The Community Epidemiology Work Group, also supported by NIDA, collects reports from drug abuse prevention and treatment experts and officials around the country about patterns and trends as they perceive them in their own areas, and summarizes and publishes those reports in volumes of "Proceedings". The local reports are not in standard format, and CEWG does not generate statistical series. Prices are occasionally mentioned; quantities rarely, if ever. It is conceivable that CEWG participants night be encouraged and funded to develop informal "panel studies" among drug users in their own areas. But this would represent a substantial move away from CEWG's current informality.

Changes and Other Studies

As suggested above, drug abuse consequence data have little promise of contributing much to making or evaluating strategy against organized crime. The one source not now tapped is arrest data. Arrests of users - particularly heroin users - for property crimes and prostitution may be quite sensitive indicators of short-run changes in availability. To develop such data, it would be necessary to test arrestees' urine for the presence of opiate metabolites and quinine. The National Institute of Justice is funding a few pilot programs along these lines. Assuming they turn out to be operationally feasible, they probably represent a wise investment of funds.

Enforcement Data

From the agencies that arrest, prosecute, try, and imprison drug traffickers and seize their drugs and other assets, we can gather valuable data both about the enforcement activity and about the underlying illegitimate market. Some of this information exists, or can readily be obtained, in statistical form; the rest is buried in case reports, informant interviews, and intelligence analyses. The one governmentally-published series estimating the revenues in the drug markets, the annual Narcotics Intelligence Estimate, is prepared primarily by enforcement agencies (particularly DEA) and apparently in large part from enforcement sources.

Information About Enforcement Activity

Three things about enforcement activity are of interest:

(1) Inputs. What resources - measured in dollars, agent-hours, prosecutor-hours, cell-years, ship-days, wiretap orders under Title III - are spent on drug enforcement, and how are those resources divided among agencies, geographically, by target drug, by enforcement technique, and by the trafficking activity under investigation (high-seas smuggling, port smuggling, first domestic distribution, wholesaling and middleman activity, retail dealing, money laundering)?

Even at the Federal level alone, something as relatively simple as a comprehensive dollar budget broken down by agency and by drug is not available. An attempt has been made to estimate the share of marijuana in total Federal enforcement; it illustrates both what is and what is not available. Assembling such data is best done within the government, by an agency with the authority to demand the relevant documents, than by outsiders; the Drug policy Board, the Office of Management and Budget, the White House Drug Abuse policy Office, any one of the cognizant Congressional subcommittees, the General Accounting Office, or the President's Commission on Organized Crime could, by collating budget submissions from the agencies involved, assemble crude but adequate estimates of dollars spent by agency and drug rather easily.

Assembling meaningful information about inputs in non-monetary forms (e.g., work-hours) would be a tedious process, requiring both access to internal agency reporting systems and an understanding of how those systems represent or misrepresent reality. However, assuming that something resembling the DEA's annual statistical report is available from the Coast Guard, Customs Service, IRS, and FBI, it might be possible to assemble a fair picture of Federal enforcement activity. Federal prosecution agencies do not now have the ability to generate statistical profiles of their work that would be useful here. The case-management software system (PROMIS) being installed in the U.S. Attorneys' offices may be able to produce estimates of prosecutive work-hours by case type, although the system may not code enough of the case data (e.g., G-DEP rank of case and defendants) to allow useful analysis.

The Administrative Office of U.S. Courts can produce cases-pending counts by offense, although this overestimates drug cases as a proportion of total cases to the extent that drug defendants are more likely than others to default on bail and thus remain as active cases for long periods of time; no estimates are made of the use of judge-days or grand jury days for various kinds of cases. Moreover, since the statute names and numbers attached to a given drug charge (manufacture, distribution, conspiracy, continuing criminal enterprise, etc.) do not vary with drug involved, it is almost impossible to separate the conviction data by drug. It is possible to make very crude estimates of Federal prison-cell utilization by convicted drug dealers from the court data. The attempt to work directly from DEA data is frustrated by the fact that the DEA statistical report lists total convictions, rather than convictions in Federal courts. The lack of a unified set of file numbers for the federal criminal-justice agencies, or a set of links from one file system to the next (what is called an "Offender-Based Tracking system," or OBTS) makes it nearly impossible to trace a sample of DEA defendants through prosecution and incarceration.

The primary problem with assembling such data is not technical, but organizational: its collection and dissemination are not in the perceived interests of the agencies involved or the key individuals within them. While that remains, little progress is to be expected. Below the Federal level, the situation is much worse; there are simply no published figures on drug enforcement spending or activities by state and local police agencies, most of which do not have separate drug enforcement units.

The current paucity of data on enforcement activities would frustrate any attempt to study the effects of changes in enforcement resources on the drug markets, even if adequate price and quantity data were in hand. This problem is not easy to fix; unless the agencies involved have their own reasons for collecting the data, it will be difficult either for "outsiders" - even from within the government - to do the job or to persuade the agencies to do it instead. The situation is similar for state and local enforcement; in the absence of any convenient ways of getting nationwide data, a few studies of individual cities and states might be useful.

(2) Outputs. How do those inputs translate into enforcement actions felt by traffickers: arrests, convictions, sentences, time actually served in prison, drug seizures, asset seizures? How are those outputs distributed geographically, by drug, by level of the traffic?

These figures are less nebulous than resource and workload data; a man-hour is "the stuff that dreams are made on," but an arrest actually happens. Nonetheless, although DEA reports its arrests broken down by almost every possible variable, there is no government-wide collation of all arrest reports. Convictions and prison time, as noted above, are simply not available in useful formats.

Even drug seizure data are complicated by the multiplicity of agencies involved; simply adding together agency totals runs the risk of double-counting. The first annual report of the Organized Crime Drug Trafficking Task Force program gives government-wide numbers that may be reliable.

Data on asset seizures are complicated both by interagency (and Federal/ State/local) cooperation and by the complexity of seizure cases: appraised value is one thing, realized value less lienholders' equity something very different. The lags in forfeiture actions against seized items, plus the growth in the overall level of seizures, make it difficult to compute seizure-to-forfeiture ratios, because 1984 forfeitures do not correspond to 1984 seizures. Some central responsibility for this area was assigned to the U.S. Marshal Service and the Asset Forfeiture Office of the Criminal Division of the Justice Department; those agencies may now be in a position to provide unified statistics.

Arrests, convictions, seizures, and prison time are basic measures of risk imposed on the illicit drug industry, and it is very hard to plan or evaluate policy in their absence. Unlike most of the input numbers, good output measures can be assembled with only minimal cooperation from the agencies themselves, if someone with authority insists on it.

Below the Federal level the situation is more dismal. Other than the uninformative counts of total drug arrests in the Uniform Crime Reports, no national data are available. It has been almost ten years since the last national study of state prisoners by offense category, and even that study combined all drug offenses into one category. Here again, the best that can be hoped for is a few well-done local studies.

(3) Effects. Enforcement actions impose costs and risks on illicit entrepreneurs and thus tend to increase prices and perhaps to make supplies less reliable. To some extent the cost-imposition value of a given enforcement action - the imprisonment of a major marijuana importer for an effective term of six years, the seizure of one kilogram of 90 percent pure heroin from a New York City wholesaler - can be calculated, and those costs can be added together to estimate the effects of enforcement on the market. Such calculations could contribute substantially to policy analysis and evaluation. They could form the basis for comparisons of alternative enforcement techniques in terms of the costs imposed on a given drug market per government dollar spent. However, to make then accurately one needs the right kind of tabulation of enforcement outputs.

The value of any illicit drug grows enormously as it moves down the distribution chain toward the final consumer. A kilogram of heroin that sells for $200,000 is the raw material for street bags of heroin worth more than $2 million. Adding together physical volumes seized at different stages of the traffic is adding apples and oranges. Each seizure should be evaluated at the point where it is seized, and then the values added together. The "street value" calculations, so beloved by reporters, resemble calculations of the value of stolen cattle based on the price of a steak dinner in a restaurant. Given estimates of the "price chains" (of which more below), simple rules of thumb, based on quantity and location, will give reasonable estimates of the coast imposed on the illicit market by drug seizures. Imprisonment data also need to be quality-weighted. The higher in the trade a trafficker is, the more dollars he would spend to avoid a year in prison, both because he has more dollars and because any given period in prison costs him more in lost earnings. That willingness-to-pay per prison-year-willingness-to-pay of traffickers at different levels. But unless the data on imprisonment are collected by drug and G-DEP violator class, there are no figures to use in calculation.

Information about Drug Trafficking and Organized Crime

Enforcement activity yields two kinds of information about drug trafficking: (1) that which can be inferred from the enforcement statistics, and (2) that known to (and sometimes published by) agents, prosecutors, and enforcement-agency intelligence analysts.

(1) Inferences from Enforcement Statistics. Lost in the organizational history of the Federal Bureau of Narcotics is the unsung hero who first proposed the rule of thumb that enforcement agencies capture about 10 percent of the drug supply on its way to market. That rule of thumb allows (once the problems with the seizure data are straightened away) a simple calculation of the physical size of the drug markets: take seizures and add a zero. More of the official drug market estimation than anyone would like to admit is still based on some variant of that methodology.

There are two problems with this calculation. First, there was never a period when we had a good measure of the total market traffickers will tend to lower it. In fact, if the change in enforcement is great enough to cause shifts in basic smuggling patterns, the result may be a lower overall seizure rate despite a higher level of enforcement. This seems to have been the case for the South Florida Task Force, which succeeded both in driving drug smugglers away from its territory and in causing many of them to switch from large vessels to smaller, faster vessels; total seizures actually fell.

The same is true of using other enforcement measures as estimates of market size. The enforcement success count depends on the size of the market, the vigor and tactics of enforcement, and the countermeasures of traffickers. Without an independent basis for estimating the effect of countermeasures, the success count tells us little even about trends in the size of the market.

(2) Facts and Analysis From Agents and Agencies. Drug enforcement agencies are not in business to collect data or do analysis; they are in business to put drug dealers in jail and seize their drugs. Like other agencies, they tend to collect and publish data that will help them do their job better, will persuade someone else to help them do their job better, or will make them look good.

Traditionally, the Drug Enforcement Agency and its predecessors defined their mission as controlling the supply of drugs. Thus, for example, the estimates of the price and purity of heroin at retail were published as part of the "Performance Measurement System," implying that drug availability was an important measure of agency effectiveness. This mission definition encouraged the development within DEA of an extensive "strategic intelligence" capability, devoted to understanding the drug markets. By contrast, the Federal Bureau of Investigation has always defined its mission as securing the conviction of lawbreakers. It regarded the number of bank robbers it caught, rather than the number of bank robberies, as its central performance measure. Since the 1981 reorganization that placed DEA under the administrative authority of the Director of the FBI, DEA's mission definition has moved closer to the FBI's. At least in part, this reflected a belief that enforcement outputs were more subject to organizational control, and thus more likely to consistently reflect organizational success, than drug market outcomes. This new mission definition may have advantages from the perspective of organized crime control. The supply-control approach tended to stress quantities of drugs rather than trafficking organizations as targets. However, the new mission definition is much less hospitable to strategic, as opposed to operational, intelligence.

(a) Price Chains. In the past DEA devoted considerable energy to estimating "price chains" for drugs, based on undercover buys and data from informants. That information, insofar as it was accurate, told a great deal about the structure of the markets, and in particular about where the money was being made. It was never formally published, but it was estimated on a quarterly basis and available to the rest of the government. It appears that such "price chain" estimates are no longer made, or no longer made available. If true, that should be rectified. There is no other source for that information, and it is essential. There is an enormous amount of relevant information in DEA case files. Undercover transactions, intercepted communications, and conversations with informants all contain price data. The problem is maintaining a staff to collate and analyze what is there. One improvement that could easily be made in publishing all statistics on drug prices is to adjust past prices for the effects of inflation, using the Consumer Price Index or GNP deflator. Otherwise, anyone trying to determine price trends has to keep looking up the rate of inflation since 1976, or whatever appropriate date.

(b) "Monitor" Programs. DEA has never had a substantial number of retail-level cases, but the Performance Measurement System required data about retail price and purities for heroin across the country. These were accordingly collected by the simple expedient of sending agents out to make undercover retail purchases and collecting retail samples from local police to seek good quality at low prices? More seriously, "retail" included samples as large as 2 grams and up to 14 percent pure, while the typical actual retail transaction is for 100 milligrams (one-tenth of a gram) less than 5 percent pure.

Now that price-purity data are no longer publicly released, there seems to be somewhat less monitoring activity than in the past. It may well be true that a program run by local police units (perhaps with funding from DEA or NIJ) would be more effective than DEA-run monitoring efforts. (Unofficial researchers might develop even more accurate results, but the legal and ethical problems seem insurmountable). But whether run by DEA or outside, there is no substitute for actual street buys as an index of market conditions, and not only for heroin.

Enforcement-based Analysis: NNICC

The National Narcotics Intelligence Consumers Committee (NNICC, pronounced "nick") is the official custodian of estimates regarding the drug markets. It was originally established to help control the embarrassing flow of mutually contradictory estimates that used to emerge from various Federal agencies. Although staffed primarily from DEA, it includes a broad variety of agencies as members, and its reports are negotiated products.

At one time, the NNICC annual report, called the Narcotics Intelligence Estimate, provided sources, physical volumes, and dollar revenues for each of the four major drug categories: opiates, cannabis, cocaine, and the "dangerous drugs." Subsequently, criticism of the dollar-volume estimates, both from outsiders and from within the NNICC staff, led to a decision to cease publication of dollar-volume estimates. At the same time the marijuana estimates for previous years were restated downward by approximately one-third. As the NNICC methodology review makes clear, the NIE figures are largely derived from estimates of farm-level production in foreign source countries, because the data sources on the consumption side are so inadequate. However, that same report shows how weakly grounded the production-based estimates are. The crop figures themselves are hard to distinguish from guesswork, and the fraction of each country's crop reaching the U.S. market is fundamentally unknown.

Still more serious is a problem not considered in the methodology review: that a large fraction of the total crop in the average year may go unsold. But a peasant marijuana-grower, for example, deciding how many plants to grow under uncertainty both about crop yields and about demand from exporters, will, if he is rational, overplant on average. Marijuana is so lucrative, compared with other crops, that it makes sense to grow an extra plant, even if there is only one chance in five of selling it for export. Of course, if there were a smoothly functioning competitive market over wide geographic areas, the result would be lower prices rather than unsold crops, but the market in marijuana for export may be highly localized, with exporters buying from nearby farmers even if prices elsewhere are much lower.

In general, then, it is hard to place much confidence in the NIE numbers, particularly because they are presented as a just set of conclusions, without any detailed presentation of the underlying data or the calculations performed on that data. One way to improve this situation would be to gather better data, particularly about consumption, as discussed above.

On the purely analytic side, two things might be done: (1) NNICC might publish more of its underlying data and some of the details about the calculations performed on them, or (2) a team of outside scholars might be allowed full access to NNICC materials and asked to prepare an independent estimate.

But whether one considers improving the data with which NNICC has to work or changing the analytic approach, the fundamental problem remains one of organizational incentives. The current lack of adequate data or convincing analysis is not a problem from the perspective of DEA or other enforcement agencies. If the information were available, there is no reason to expect that it would be used to reshape what agents and prosecutors do. In fact, by providing outsiders with a basis on which to criticize, data and analysis might seem to threaten agency autonomy. Nor have the White House or Congressional staffs concerned with these problems displayed any anxiety over the absence of information. The first step in solving a problem is getting it identified as a problem by someone with the resources to bring about a solution. In this area, that step remains to be taken.