2. Drug Supply Estimates

This section discusses the information and assumptions we used to estimate the supply of cocaine and heroin to the United States. For reasons discussed below, it is not practical to develop estimates for marijuana, methamphetamine, or other illegal drugs.

Cocaine

The process for estimating cocaine supply has been evolving over the past ten years. Since 1990, ONDCP has estimated the supply of cocaine by beginning with the potential cocaine production estimate and sequentially decreasing this amount by subtracting losses. The potential cocaine production estimate was based on imagery of coca crop fields, whose figures were then combined with leaf yield, alkaloid content, and base processing efficiency multipliers. In 1996, a U.S.-intelligence working group initiated an event-based process for estimating the amount and routes of cocaine departing South America. In March 2000, the Crime and Narcotic Center integrated data on potential cocaine production estimates with Western-hemisphere consumption estimates to calculate the amount of cocaine available for the non-U.S. markets. Our approach was to design a cocaine flow model, which standardized the terms and measures, so various existing estimation-processes (e.g., coca cultivation, domestic consumption estimates) could be integrated into one complete and coherent set of flow estimates. This model attempts to triangulate a coherent estimate of cocaine availability along the entire route of cocaine flow, and is referred to as the Sequential Transition and Reduction (STAR) model.

The STAR model incorporates diverse estimates of the production and distribution of cocaine into one cohesive, connected model. The model hinges on the notion of a transition, or movement, of cocaine from one stage in the production/distribution process to the next stage in that process. A transition is a computational link between stages that, after accounting for reductions (seizures, losses, etc.), converts drug (or drug precursor) at one stage into drug at another stage. Stages are geographic locations corresponding to established levels (e.g., political borders, growing areas, transshipment countries) in the course of drug (or source constituent) flowing from source to street. Details regarding this model are available in a companion report.⁴⁶ Readers should consult that report for specifics; a summary follows.

Description of Stages

This model establishes a coherent set of stages, established on the basis of existing supply-reduction strategies, which conform to the trafficker's patterns in cultivation, production, transshipment, and distribution. Mathematically, the model links supply estimates at each stage by transition matrices that account for conversions in cocaine state, reductions such as consumption and seizures, and geographic routing of the cocaine. In this way, the model contains a consistency between the "micro" flow within a geographic region and the "macro" estimates of cocaine supply between stages. Figure 1 presents a geographic presentation of the nine stages of movement describing cocaine supply.

Figure 1—Description of stages in the flow of cocaine from source to street

Click for larger image

Each of the stages can be described as follows:

• Stage 1, Net coca cultivation for the previous year. Expressed in hectares and is distributed among the various coca-growing areas of the Andean Ridge.
• Stage 2, Net coca cultivation for current year. Expressed in hectares and is calculated from taking the previous stage and accounting for new growth and reductions from eradication and field abandonment in the various growing areas.
• Stage 3, Net leaf tonnage. Expressed in metric tons and is determined by applying leaf-yield conversions to the previous stage, then accounting for leaf seizure and consumption reductions.
• Stage 4, Cocaine base: is expressed in metric tons of cocaine base and is determined by applying alkaloid-content and lab processing efficiency figures to the previous stage and accounting for cocaine-base seizure reductions.
• Stage 5, Cocaine at HCl labs. Expressed in metric tons of cocaine, is measured at the HCl labs distributed within South America, and accounts for losses of cocaine-HCl at the labs.
• Stage 6, South American departure areas. Expressed in metric tons of cocaine, is measured at the South American departure areas, and is reduced by South American seizure and consumption losses.
• Stages 7a and b, Transshipment area and world markets. After departure from South America, cocaine is smuggled toward its markets in the United States, Canada, Europe, and the rest of the world. Most of the cocaine destined for the United States is initially smuggled to transshipment locations (Stage 7a) in Mexico, Central America, and the Caribbean islands including the Bahamas and the Antilles. Additionally, cocaine is shipped to non-U.S./Latin American markets overseas and in Canada (Stage 7b). Cocaine estimates at both stages 7a and 7b are reduced by en-route losses due to en-route seizures and consumption in the transshipment countries.
• Stage 8, U.S. border. From the transshipment areas, cocaine moves across the U.S. border, after accounting for seizure losses at the U.S. border.
• Stage 9, U.S. retail locations. From the border, cocaine is transported to retail markets in the United States, after accounting for domestic seizures.

Although the STAR Model theoretically provides a complete and coherent set of connected stages, input data was not always available. Additionally, supply data has varying degrees of certainty. As a result, the STAR Model combines data from various sources to triangulate an estimate of cocaine availability. One of the triangulation legs begins with the coca cultivation estimates and works toward an annual estimate of cocaine available for export from South America. The second leg in the triangulation begins with the domestic consumption estimate, described earlier in this paper, and works backward toward an independent estimate of cocaine departing South America. The third triangulation leg is the event-based estimate of cocaine availability developed by the Interagency Assessment of Cocaine Movement (IACM) working group, which also estimates the amount of cocaine annually departing South America. Development of each of these triangulation estimates will be described, and then compared to illustrate the coherence of the STAR Model estimates.

Cultivation-based Supply Estimates

The STAR Model starts with data on cultivation and cocaine processing. CNC uses statistical survey methods, similar to those employed by agricultural organizations estimating the size of licit crops, to estimate the quantity of coca under cultivation in Colombia, Peru and Bolivia. CNC's survey randomly samples potential growing areas, placing a higher sampling probability on known growing regions, and satellites and airplanes then photograph the selected areas. CNC analysts interpret the resulting images to develop country-wide coca crop estimates. The uncertainly in this approach has been estimated by CNC to be +/-10%.

Operation Breakthrough, a series of studies done by the DEA, provides data on coca crop productivity and base processing efficiencies. The three critical factors in calculating cocaine production from the cultivation estimates are the leaf yields, alkaloid content of the coca leaf, and the base processing efficiency. These factors can have significant uncertainty during transition periods, such as 1993–99, when Colombian cultivation increased dramatically. Figure 2 depicts the annual changes in the distribution of Andean potential production, and the effect of the revised estimates.

Figure 2—Andean Potential Cocaine Production Estimates, 1990–1999 (pure metric tons)

Click for larger image

Table 12 shows the STAR Model's estimates of cocaine and its source-constituents, from cultivation, through production, to export from South America. The reader should be aware that these figures will be lower than the annual potential production estimates because they account for losses such as leaf seizures and spoilage, base seizures, and HCl seizures in South America. The STAR Model estimates for cocaine at the various stages is discontinuous from Stage 5 (at the HCl labs) to Stage 6 (at the South American departure areas) because that transition requires an estimation of South American cocaine consumption, which is currently not available. Stage 5+, shown in Table 12 below, represents the estimate of cocaine supply available for export (or consumption in South America), once South American seizures have been subtracted from the Stage 5 estimate.

Table 12—Net cocaine produced for illicit markets (units as noted)

Click for larger image

Domestic Consumption-based Supply Estimates

Once the stages and transitions were established by the STAR Model, cocaine supply estimates could be calculated by starting at either end: either by beginning with the coca cultivation estimates and working forward to estimate cocaine supply available for domestic consumption, or by beginning with the domestic consumption estimates and working backward to estimate actual cocaine production. This section will describe the latter approach.

An estimate of cocaine availability at departure from South America (Stage 6) was determined by the STAR Model, based on the domestic consumption estimate, discussed earlier. The annual estimates of domestic cocaine consumption, shown in Table 7, were input into the STAR Model for Stage 9. From this estimate, losses such as domestic, border and transshipment seizures were added. Consumption estimates for non-U.S. countries were also added to the domestic consumption figures to result in an estimate of actual cocaine production departing South America. Table 13 shows the stage-by-stage figures.

Table 13—Net cocaine produced for domestic retail market (metric tons)

Click for larger image

Event-based Supply Estimates

The Interagency Assessment of Cocaine Movement (IACM)ⁱⁱ uses an event-based, interagency consensus methodology to quantify intelligence reports about cocaine movement through the transit zone. Each quarter, intelligence and operations analysts from the various interdiction agencies meet to discuss their perception of cocaine movements departing South America. If the information for a particular event is sufficient, the event is included into a data base, and pertinent data on the event is recorded. One piece of information is the load-size of the cocaine contraband conveyed by the movement. This load-size is based on one of three sources: observation (usually a seizure), confidential informants, or historical trend analysis. Once the data base of events is complete for the year, the sum of the load-sizes provides an estimate of cocaine departing South America.

Table 14 shows the annual estimates for cocaine flow through each transshipment corridor, assuming export quality purity. The annual total is converted to pure metric tons by adjusting for purity, to result in an independent estimate of cocaine departing South America.

Table 14—Event-Based Cocaine Amounts Departing South America By Transit Corridor, 1996–1999 (bulk metric tons)ⁱⁱⁱ

Click for larger image

Comparison of Cocaine Supply Estimates

The three sets of figures described above all provide estimates of cocaine departing South America. Figure 3 compares these three annual estimates of cocaine departing South America:

1)	based on the coca cultivation estimates, from the STAR Model,
2)	based on the domestic consumption estimates, from the STAR Model, and
3)	based on an assessment of movement events.

The domestic consumption-based and the event-based estimates correlate closely in magnitude (500–600 mt/year), and in trend. An uncertainty bar of -148 metric tons was attached to the cultivation-based estimate to account for the unknown South American consumption losses. This was the estimate of South American consumption for 1999 developed by the Crime and Narcotics Center. When the uncertainty-bars are included, the cultivation based estimate encompasses the other two figures for 1998–99. The STAR estimate of cocaine departing South America shows a decreasing trend over the four years, which is not consistent with other trends. Worldwide seizures and domestic consumption have been stable over the past four years; Latin American and European consumption is believed to be increasing; therefore, cocaine availability for world consumption should be stable or increasing. But without better cocaine cultivation and production data, these uncertainties will remain.

Figure 3—Comparison of cocaine availability estimates, metric tons

Click for larger image

For this paper, we wanted to compare the supply estimates with our consumption estimates to understand the reasonableness of our approach. To this end, we used the STAR Model to extrapolate both the cultivation-based estimate and the event-based estimates of cocaine departing South America, to calculate domestic consumption. Each estimate of cocaine availability departing South America was reduced by the figures shown in Table 15 below.

Table 15—Cocaine Losses (pure metric tons)

Click for larger image

Table 16 below shows a comparison of domestic consumption estimates based on three approaches: 1) consumption estimate explained in this paper, 2) the event-based estimate, and the cultivation-based estimate. The cultivation-based estimate is shown as a range, because the 148 metric ton estimate of South American consumption has been subtracted for the lower limit.

Table 16—Comparison of Domestic Consumption Estimates (pure metric tons)

Click for larger image

Heroin

The modeling approach used for heroin differs from that for cocaine. While the bulk of cocaine production is destined for the United States, less than five percent of worldwide heroin/opiate production is sent to the United States, so modeling the flow from production to consumption is impractical. Also, dissimilar data are collected for heroin and cocaine. For example, heroin has no counterpart to the Interagency Assessment of Cocaine Movement (IACM), so we know less about the dynamics of heroin movement than about cocaine movement. On the other hand, cocaine has no counterpart to the DEA's Domestic Monitor Program (DMP) and Heroin Signature Program (HSP). A heroin availability model must differ from a cocaine availability model, because it is constructed from a different empirical base.

This section presents a model of the movement of heroin into the United States. Details appear in a companion report.⁴⁷ We do not consider the model as final, because data about heroin trafficking continues to grow, and modeling improvements will follow from better data. Nevertheless, the model is an important step toward structuring what is currently known about the ways that heroin suppliers provide drugs to the United States. Like its cocaine counterpart model, the heroin flow model seeks to weave together and reconcile various estimation systems into one comprehensive model.

Model of Heroin Availability

Figure 4 depicts an overview of the heroin model. The rest of this report elaborates, and the companion report provides details. Whereas the cocaine movement model takes potential production estimates as its starting point, the heroin model begins at the other end—with the U.S. consumption estimates that were developed earlier in this report.

The source of heroin consumed in the U.S. is partitioned into four production areas: South America, Mexico, Southeast Asia and Southwest Asia. That partitioning is based on an analysis of data from the Heroin Signature and Domestic Monitor Programs, first done by Abt Associates for the Drug Enforcement Administration⁴⁸ and later extended for the Office of National Drug Control Policy.

The Federal-Wide Drug Seizure system provides the best estimates of where heroin enters the United States. As shown subsequently, most seizures were in California, Texas (and Arizona), Florida (and Puerto Rico), and New York (including New Jersey) so the figure identifies those four principal entry points. The source country of those seizures is estimated from the Heroin Signature Program (HSP).

Figure 4—Overview of a Heroin Flow Model

The model takes into account seizures and non-U.S. consumption of South American and Mexican heroin. However, according to reports by the Community Epidemiological Working Group (CEWG) and the U.N. World Drug Report, consumption seems minimal within Colombia and Mexico, so most South American and Mexican heroin is probably destined for the United States. Because non-U.S. consumption accounts for so much of the Southeast and Southwest Asian heroin, the model accounts for heroin movement from Southeast and Southwest Asia at the U.S. border, but not earlier.

The model provides a consumption-based estimate of the amount of heroin produced in South America and Mexico. CNC provides a production-based estimate of the heroin production potential in the same areas. After accounting for seizures and other leakage, the supply-based estimates should agree with the consumption-based estimate at least roughly—if not, something is wrong with the consumption model, with CNC's production estimates, or both. CNC also estimates potential production for Southeast and Southwest Asia, but there is no apparent way to tie a consumption-based model into those estimates.

Determination of Source Area

The Drug Enforcement Administration supports two programs—the Heroin Signature Program and the Domestic Monitor Program—to determine the source area (South America, Mexico, Southeast Asia and Southwest Asia) of heroin collected at three points: seizures at ports of entry, a random sample of other seizures and purchases, and DMP purchases. We included all specimens weighing less than one gram in a retail-level sample, comprising all the DMP data and several purchases from the random sample. We used that retail-level sample to estimate the sources of heroin used in the United States.

Our inferences are based on the retail-level sample, rather than an importation-level sample, because the retail-level sample comes closest to representing heroin actually consumed in the United States. Still, raw data tabulations are not very useful, for two reasons. First, some of the retail level specimens have too little drug to afford a signature, so the source area is unknown. This creates some problems, because Mexican heroin is easily identified and therefore is rarely classified as unknown. To prevent Mexican heroin from being over-represented in the data, we developed imputation routines for assigning a signature to every sample in the retail level data where an imputation seemed justified. Second, the Domestic Monitor Program oversamples in places where heroin use is relatively rare. (For example, St. Louis has a quarterly sample size of 10 purchases, while Baltimore has the same sample size but many more heroin users and purchases.) We developed a weighting procedure so that the signature program would represent a national estimate.

We have been unable to classify about 10% of the heroin seized and purchased since 1995. These unclassified samples are reported as unknown (UNK) in Table 17, which details estimates for the percentage of heroin from each source area. Because data were not available for 1998 and later, the 1998 and 1999 estimates are projections—that is, they are the averages for 1995 through 1997.

If we are correct about these percentages, and if we are correct that between 1995 and 1998 about 12 to 13 metric tons of heroin used per year in the United States, then we can derive estimates of the amount of heroin that come from each area (Table 18). We do not provide estimates before 1995, because the unknown signature category is comparatively large before 1995.

Table 17—Source of Heroin Used in the United States (Projected for 1998 and 1999) (Percentages)

Click for larger image

According to these calculations, U.S. consumers use somewhat less than 7 metric tons of South American heroin and somewhat more than 3 metric tons of Mexican heroin. However, the South American and the Southeast and Southwest Asian numbers might be somewhat higher depending on how the unknown signatures are partitioned across the data.

Table 18—Estimated Amount of Heroin from Each Source Area (metric tons)

Click for larger image

Seizure Levels

Some foreign production gets seized as it enters the United States. We tabulated heroin seizures reported in the FDSS from 1991 through the first half of 1998. To provide greater comparability between 1998 and earlier years, we interpolated seizures for the entire year by doubling seizures from the first half of 1998. The figure seems to show that seizures have varied between about 1.2 and 1.6 metric tons from 1991 through 1998. There is no apparent trend.

There is a second useful way to look at these data. Between 1991 and 1998, 99.2 percent of all seizures were less than 10 kilograms. Likewise, 99.7 percent of all seizures were less than 20 kilograms and 99.9 percent of all seizures were less than 50 kilograms. If we exclude all seizures larger than 50 kilograms from the tabulation, seizures have remained fairly constant at about 1.2 metric tons. Apparently, exceptionally large seizures can occasionally lead to spikes in the seizures observed during any year, distorting the trend. When large seizures are included in the estimates, an annual seizure rate of 1.3 metric tons seems representative of law enforcement success at preventing heroin from entering the United States.

Figure 5—Heroin Seized by Year Metric Tons

Click for larger image

In fact, when imported into the United States, heroin is typically about 80 percent pure.⁴⁹ Thus the 1.3 metric tons of bulk heroin probably translate into somewhat more than 1 metric ton of pure heroin being seized as it enters the United States. According to the 1999 INCSR, Mexican authorities have seized between 0.14 and 0.38 metric tons of heroin (or opium equivalent) every year since 1995. Given what U.S. authorities seize, Mexican traffickers seem to lose on average about 0.34 metric tons per year. Colombian authorities never seized more than about 0.15 metric tons per year, so seizures probably account for an average of about 0.75 metric tons of Colombia's production per year.

Importation Points

Where do these seizures occur? Most seizures happen in one of four importation areas, defined:

• New York (includes New Jersey)
• Florida (includes Puerto Rico)
• California
• Texas (includes Arizona)

The rest of the seizures occur throughout the United States.

Figure 6—Proportion of Heroin Seized by State (Region) Weighted by Seizure Size

Click for larger image

The curves shown in Figure 6 are a smoothed representation of how the location of seizures changed over time. Seizures have been weighted to reflect the amount of heroin involved in the shipment. A companion report explains the methodology used to develop these curves.⁵⁰

The figure shows that the proportion of seizures made in New York, represented by the highest line in this figure, decreased precipitously from 1991 through 1995 and then stabilized. Most of that reduction was balanced by a dramatic increase and then stabilization of seizures made in Florida. The figure suggests that more heroin was being shipped to New York during 1998 than was the case in 1996 and 1997. This may be true, or it may be that a few especially large shipments have distorted the trend. Also, the smoothing procedure can distort trends at the end of the period. It would be prudent, therefore, to discount the apparent increase in New York seizures and decrease in Florida seizures observed in 1998.

One point is clear: By 1995, seizures had decreased markedly in New York, and they had increased correspondingly in Florida. There was little change in seizures in the rest of the nation. Using the geography of seizures as an indication, after 1995 the geographic movement of heroin into the United States has been relatively stable.

Movement of Heroin from Source Areas into the United States

Table 19 reports the source of heroin that was seized in the five areas identified in the previous figure. This table is based on seizures made at airports, at the borders, and through the mail. The probability that a shipment is seized likely varies across conveyance mode and geographic location, so a simple tabulation of seizure data would be a biased representation of where heroin enters the United States. To make the tabulations more representative of heroin imports, we weighted the data so that the source area of heroin seized was the same percentage as the source area of heroin used in the United States.⁵¹ Estimates of the source areas of heroin in the United States have been reported already in Table 18.

Table 19 should be read down its columns. For example, an estimated 82 percent of the heroin that entered the U.S. through California came from Mexico. Almost 86 percent of the heroin that entered through Florida came from South America.

Table 19—Estimated Percentage of Heroin Entering the United States by Importation Point for Each Source Area

Click for larger image

Table 20 reports the estimated percentage of heroin from each source region that entered the United States through each of the five importation areas. This table should be read across its rows.

Table 20—Estimated Percentage of Heroin Entering the United States by Source Area for Each Importation Point

Click for larger image

If weighted seizures are a good reflection of where heroin enters the United States, then 64.3 percent of Mexican heroin enters through California and 16.3 percent enters through Texas. That is, more than 80 percent of Mexican heroin probably comes across the Southwest border, and the rest of Mexican heroin enters the United States through other diverse locations. More than half of South American heroin enters the United States through Florida, and most of the rest comes through New York. Almost three-quarters of Southeast Asian heroin enters through New York and the rest goes through diverse places. Three-quarters of the Southwest Asian heroin also seems to enter through New York City, and the rest goes through various places. The increased role of South America as a supplier of heroin explains why Florida has become an increasingly important heroin importation point.

Table 20 provides another useful way to summarize these data. Multiplying the percentages by source area and importation point (Table 18) by the amounts per source area (Table 16) provides an estimate of metric tons moved through each importation point by source area. To develop this estimate, we average across the five years reported in Table 16.

If we are correct that Americans used about 12.3 metric tons of heroin per year between 1995 and 1998, then Table 21 gives some idea of how much heroin from each source moves into the country through each region of the United States. Of course, there exists considerable uncertainty in estimates that provide this much detail.

Table 21—Estimated Amount of Heroin (Metric Tons) Entering the United States by Source Area and Importation Point, 1995–1998

Click for larger image

Almost 10 percent of the heroin was classified as unknown—that is, DEA chemists could not assign a source area to that heroin. Note that, excluding the unknown category, virtually all heroin seized in Florida came from South America. It seems reasonable to suppose that most of the 13.5 percent of the heroin seized in Florida and identified as "unknown" also came from South America. This same reasoning cannot be applied to other places where South America is not the dominant supplier, but it does suggest that South America's share of the U.S. market may be greater than is indicated by Tables 16 and 19.

CNC Potential Production Estimates

How do our estimates of the amount of heroin from the producer nations compare with CNC's reports of production potential? Since 1995, CNC has consistently estimated the production potential of South America at about 6.1 to 7.5 metric tons. (These estimates are after subtracting eradication losses from total hectares. The 7.5 metric ton figure is for 1999; it was never previously larger than 6.6 metric tons.) Unfortunately, estimates are of uncertain accuracy because the assumed conversion ratios from poppy to opium is based on intelligence fieldwork in Southeast and Southwest Asia. We cannot know for sure whether or not those conversions apply to South America. Nevertheless, we must take those conversion estimates as the best currently available.

According to our consumption estimates, Americans consume somewhat more than 6 metric tons of heroin from South America, and United States authorities seize about 0.75 metric tons. Our consumption/seizure estimates exceed South America's production capacity, but the difference is not great. This suggests that the estimated 12 to 13 metric tons of total domestic heroin consumption is about right if somewhat high.

Since 1995, CNC's estimates of the production potential for Mexico vary over time between 4.3 and 6.0 metric tons. According to our estimates, Americans consume somewhat more than 3 metric tons of Mexican heroin and another 0.34 metric tons are seized by U.S. or Mexican authorities. The consumption-based estimates are less than the production-based estimates. The Mexican production estimates suggest that the estimated 12 to 13 metric tons of domestic heroin consumption is too low.

CNC's production estimates for Mexico are inconsistent with our consumption estimates. There seems to be no ready reconciliation, but speculation may be helpful. CNC emphasizes that its estimates are for potential production, and actual production may differ. Perhaps Mexico's production is well below its potential, but it is difficult to reason why potential production would be consistently less than realized production. A better explanation comes from CNC's warning that:

The wide variation in processing efficiency achieved by traffickers complicates the task of estimating the quantity of cocaine or heroin that could be refined from a crop. These variations occur because of differences in the origin and quality of the raw material used, the technical processing method employed, the size and sophistication of laboratories, the experience of local workers and chemists, and decisions made in response to enforcement pressures. (INCSR, 1999)

CNC's assumptions may overstate Mexico's production efficiency. This is speculation, of course, but we observe that heroin imports are about 44 percent pure when from Mexico, 80 percent pure when from Colombia, and 70 to 75 percent pure when from Southeast and Southwest Asia. Because CNC makes the same assumptions about production efficiency for Mexico as it does for Southeast and Southwest Asia, the potential production may overstate Mexico's actual production.

Suppose that Mexican production were 0.59 as efficient as is assumed by CNC. (The 0.59 comes from dividing 0.44 purity by 0.75 purity.) Then an estimate of Mexico's actual production would be between 2.5 and 3.5 metric tons, numbers that agree with the consumption estimates. Using this same argument, we might assert that Colombian production is 1.07 times more efficient than is assumed by CNC. This would lead to a higher estimate of Colombia's production, which would be more consistent with the consumption estimates. This reasoning is speculative, but not unreasonable in the face of having no reliable data about the actual production efficiency in Mexico and Colombia.

The intelligence community has estimated that, during the late 1990s, Americans used about 18 metric tons of heroin per year. To get this estimate, the community accepted the ONDCP estimate of 980,000 hardcore heroin users and assumed those users consumed an average of 50 mg per day. Use by occasional users was apparently factored into these calculations, but the method is unclear.

This amount is considerably more than the 12 to 13 metric tons estimated in this report. The intelligence community considers the 50 mg per day estimate to be conservative. Indeed, some addicts can use much more as evidenced by consumption by opiate users who enter treatment. But beyond this upper bound, the 50 mg estimate seems to have no justification beyond the assertion that "Many analysts and treatment professionals, however, believe that 50 mg as the estimate for average daily dosage for heroin users in the United States underestimates overall US market demand." Thus, the 18 metric ton estimate would seem to rest on a shaky and unverifiable assumption.

This is not to say that the estimate from the intelligence community is wrong, of course. Nevertheless, if we accept the estimate of 18 metric tons, we have to deal with some inconsistencies. Perhaps those inconsistencies are ultimately resolvable, but surely they cannot be readily dismissed. For example, if we are correct that a milligram of heroin costs roughly $1, then the implied $350 per week expenditure exceeds our estimates of expenditures by hardcore users. As another example, the estimates imply that 8 metric tons of heroin come from Colombia and 5 to 6 metric tons come from Mexico. For reasons explained earlier, we doubt that Colombia can provide this amount of heroin after accounting for seizures. Furthermore, even this high estimate of 8 metric tons is lower proportionately than Colombia's apparent share of the heroin market. Mexico might be able to supply this level, presuming production estimates are realistic, but for reasons stated, we think that Mexico's production is overstated.

Non-U.S. Consumption

How much heroin is consumed within Mexico and within South America? What other reductions occur in the production and distribution systems? Unfortunately the answers to these questions are all but unknown.

Perhaps the most useful published information about consumption comes from reports of the Community Epidemiological Working Group (CEWG). The CEWG is focused on the United States, of course, but most of its reports include sections on consumption in other nations. These reports are seldom quantitative, because nations outside the United States rarely have data collection systems affording estimates of domestic consumption. Based on CEWG assessments, we assume that the consumption of heroin within South and Central America is negligible. Most heroin produced in South and Central America is probably destined for North American markets.

Canada is a bigger problem. According to CEWG reports, heroin is seen as a major drug problem, at least in Vancouver and Toronto. But we do not know the amount of heroin used in Canada; nor do we know the source.⁵² It seems reasonable to assume that some South American and Mexican heroin is shipped to Canada, but we do not yet have an estimate of the amount.

Heroin—the Supply-Side Assessment

Our best estimate is that roughly 12 to 13 metric tons of heroin is used in the United States during a given year. The level of use could be lower, of course, but if it were much lower than 12 metric tons, then we could not account for production potential in Colombia and Mexico, most of which is presumably exported to the United States. Likewise, the level could be higher, and while Mexico could be providing more than 4 metric tons, estimates of more than 12–13 metric tons would be difficult to reconcile with Colombia's apparent production capacity.

Modeling the Flow of Methamphetamines

In 1990, Mexican organized crime groups began large-scale production of methamphetamine and rapidly expanded distribution into California and other parts of the Southwest. In addition to combating large-scale production, United States government efforts to control the distribution of methamphetamines have become increasingly difficult due to the proliferation of small clandestine labs, each of which produces small quantities of the drug. Methamphetamines can be produced easily and inexpensively using chemicals bought at local drug stores or chemical supply companies. A person with little technical training can easily learn how to make methamphetamines. This has become increasingly possible due to several Internet sites that include detailed step-by-step "cooking" directions.⁵³

Prior to 1989, methamphetamines were produced primarily by outlaw motorcycle gangs using a technique called "Phenyl-2-Propnanone (P2P) synthesis." During this time, P2P was a controlled substance; however, the precursors required to make P2P were not controlled, which enabled the motorcycle gangs to "legally" produce methamphetamines. The precursors of P2P were subsequently controlled by the first US chemical control act, the 1989 Chemical Diversion Trafficking Act (CDTA).⁵⁴ After 1989, the primary methamphetamines precursor shifted from P2P to ephedrine. The ephedrine reduction method became the primary method of synthesis due to a CDTA loophole: The CDTA restricted the importation of bulk ephedrine but made no restrictions on the tablet form of the chemical.⁵⁵

From 1990 to 1994, ephedrine-based production, based in Mexico and California, was the predominant production method. During this time, methamphetamine production rapidly expanded from Mexico and the Southwest corner of the United States into the Midwest and the South.⁵⁶ The Mexican drug cartels used existing marijuana and heroin distribution networks to distribute the methamphetamines. Passage of the Domestic Chemical Diversion Control Act (DCDCA) in 1994 made ephedrine tablets a List 1 chemical, restricting their sale. This Act did not stop the Mexicans, who in 1994 began the illegal smuggling of ephedrine. Mexican drug rings purchased large amounts of ephedrine indirectly from rouge companies outside of Mexico that, in turn, purchased the chemicals and then delivered them to Mexico.⁵⁷

The DCDCA also caused a shift in methamphetamine's mode of production. Although the DCDCA controlled the sale of ephedrine, it did not control the sale of pseudoephedrine, which became the precursor of choice.⁵⁸ Pseudoephedrine is found in Sudafed and other similar over-the-counter cold medicines. This made it much easier for average criminals to get access, leading to a rapid increase in the number of small clandestine labs, especially in the Midwest. From 1994 to 1996 the number of pseudoephedrine imports into the United States (in metric tons) increased by almost 50 percent.

Clandestine labs in the Midwest primarily use a method of synthesis called the "Nazi method," because it was first used in Germany during World War II. The Nazi method has become the dominant production procedure in the Midwest because it requires ammonia, which is used throughout the Midwest in fertilizers. Stolen ammonia is the primary source of ammonia for the clandestine labs. The Nazi method is popular because it can produce a highly pure methamphetamine product very quickly: in about 3 hours, compared with the ephedrine reduction method, which can take several days. Small clandestine labs are often mobile and typically produce between 1 and 4 ounces of methamphetamine at a time.⁵⁹ From 1995 to 1996, the DEA reported a 169 percent increase in the number of DEA clandestine lab seizures (327 and 879 respectively). This trend continued in both 1997 and 1998.⁶⁰

Although the number of small clandestine labs have grown rapidly, the methamphetamine seized from them (see figure 7) only accounts for a small portion of the methamphetamine seized by the DEA from labs. In 1998 small clandestine labs accounted for 95 percent of the lab seizures, but only 22 percent of the lab-seized methamphetamine; a majority of the seized methamphetamine (78 percent) came from seizures of the super labs.⁶¹

Figure 7—Methamphetamine Clandestine Lab Seizures by DEA

Click for larger image

Modeling the flow of methamphetamines poses unique challenges. A cocaine model can begin with estimated production in known growing areas, but methamphetamine production has no comparable geographic boundaries. A heroin model can begin with consumption-by-production region estimates, but developing signatures has proved to be much more difficult with methamphetamine, primarily because of the large number of clandestine labs that have spread all over the United States. In order to develop a signature for a drug, there must be large geographic variability between different drug sites. Clandestine labs are now in almost every state in the United States, making it much more difficult to decipher between different drug sources. In addition, methamphetamine is completely synthetic. Using the Nazi method, clandestine labs can make a highly pure drug product, mitigating the levels of impurities that are necessary to accurately determine the signature of a drug. Unlike heroin or cocaine, which are grown in specific geographic locations (Columbia, Thailand, etc.), anyone can manufacture methamphetamines with the proper ingredients and cooking instructions. This adds a dimension of difficulty to finding an accurate model of methamphetamine production and distribution in the United States.

An alternate way to model the production and distribution of methamphetamine is to monitor the production and distribution of precursor chemicals. This approach has serious limitations, including the need to make allowance for the legitimate use of those precursors. For example, methamphetamine production requires a large quantity of pseudoephedrine. In order to produce 1 ounce of methamphetamine, a small lab requires 680 60 mg tables (roughly 1.44 ounces) of pseudoephedrine (based on a 70 percent conversion rate). Figure 8 shows that pseudoephedrine imports increased by roughly 200 metric tons after 1994, although this increase was only about 100 metric tons by the late 1990s. If we assume this 100 metric ton increase reflects methamphetamine production, then it represents 70 metric tons of methamphetamine. Given a typical street purity of about 40 percent, this represents just under 30 metric tons of pure methamphetamine—considerably more than the consumption-based calculations, and this does not account for production imported into the United States.

Figure 8—Ephedrine & Pseudoephedrine Imports into the United States

Click for larger image

According to the DEA,⁶² between June 1993 and December 1994, an estimated 170 metric tons of ephedrine were supplied to Mexican traffickers. Also according to the DEA, this could have yielded 170 tons of methamphetamine. Again assuming 40 percent purity, this represents almost 70 metric tons of pure methamphetamine, far in excess of the consumption-based estimates.

The above arguments are not intended to argue that the consumption-based estimates are correct while these supply-based estimates are wrong. Rather, the point is that supply-based estimates, which are based on precursor chemicals, provide estimates that are difficult to reconcile with reasonable inferences about the use of methamphetamine. According to the DEA, the 170 tons of methamphetamine were "...enough to supply 12.4 million abusers with three 10-milligram doses a day for 365 days per year." Even assuming this eighteen-month estimate implies just over 8 million hardcore methamphetamine users, DEA's estimate seems much too high. The consumption-based estimate is about 400,000 hardcore users. The NHSDA estimates about 800,000 past month users of any amphetamine during this same period, and not all these used methamphetamine. Furthermore, TEDS reports 53,000 treatment admissions in 1997, a figure than has grown from only 15,000 in 1992. It is difficult to see how 8 million daily methamphetamine users could generate only 53,000 treatment admissions, when an estimated 3.5 million weekly cocaine users generate 255,000 treatment admissions. Modeling based on precursor chemicals does not seem to provide a suitable way of estimating the supply of methamphetamine to the United States.

Marijuana

It is also difficult to develop an estimate of the size of the U.S. retail market for marijuana from estimates of available supply. First, the amount of marijuana that Americans cultivate for personal use cannot currently be estimated. Second, even though a large amount of the domestic marijuana market is grown in the United States, countries in South and Central America, the Caribbean, Asia, North Africa, and the Middle East also supply cannabis to the domestic market.⁶³ Unfortunately, the data needed to develop better estimates are not available, and, therefore, we cannot develop a plausible supply-based estimate of the retail value of the marijuana market in the United States.

Legitimately Manufactured Controlled Substances and Illicitly Manufactured Dangerous Drugs

It is impossible to know the amount of controlled substances, such as inhalants and hallucinogens, that are produced legally but diverted for illicit consumption. It is also impossible to know the amount of drugs that are manufactured illicitly in domestic or foreign laboratories. We do know that these substances are readily available.⁶⁴

Price and Purity of Illicit Drugs

Drug prices and purity offer some information about the availability of drugs in the United States. By themselves, trends in illicit drug prices are not a convincing indication of whether the demand or the supply for illicit drugs is either increasing or decreasing. For example, price might remain about the same if both the supply and the demand for drugs were increasing, but then again, a decrease in both the supply and the demand could also result in stable prices. Nevertheless, to the extent that price trends are not inconsistent with trends in supply and demand, they provide some confirmation for consumption-based and supply-based estimates.

Because illicit drugs can be bought and sold in different amounts, degrees of purity, and levels of distribution, prices can vary greatly from sale to sale. Using the Drug Enforcement Administration's System To Retrieve Information from Drug Evidence (STRIDE) data from January 1981 through June 1998,⁶⁵ we have developed statistical models to estimate typical prices for standardized purchases of cocaine, heroin, methamphetamine, and marijuana. A standardized purchase involves a set quantity and quality of drugs exchanged at a specified distribution level. A useful application of these estimates is to examine price trends for these standardized purchases over time.

• Figure 9 shows the estimated retail level⁶⁶ and importation level⁶⁷ prices per pure gram of cocaine over time. The average price per pure gram at the retail level has decreased considerably from just over $400 per pure gram in 1981 to about $170 per pure gram in 1998. The average price at the importation level has also decreased from roughly $75 per pure gram in the early 1980s to about $25 per gram in the late 1990s.

• Figure 10 compares the estimated retail-level purchase price with the estimated importation⁶⁸ purchase price of heroin. The figure shows two retail prices because the retail heroin market appears to be bifurcated into a sector selling relatively low purity heroin to injection drug users⁶⁹ and a sector selling comparatively high purity heroin to those who either inject or sniff the drug.⁷⁰ At the lowest retail level, heroin prices have fallen from about $3,000 per pure gram in 1981 to about $2,000 per pure gram in 1998. At the second retail distribution level, prices have fallen from about $2,000 per pure gram in 1981 to about $400 per pure gram in 1998. In 1998, a weighted average of the two lowest distribution levels suggests a price of roughly $1,000 per pure gram. Prices at the importation level have also fallen—from $400 to $500 per gram in the early 1980s to under $200 per pure gram in the late 1990s. In fact, border prices are probably lower, but these trends are descriptive.

• The street price⁷¹ of methamphetamine has fallen over the last twenty years (see Figure 11). In the early 1980s, prices were close to $300 per pure gram. By the late 1990s, methamphetamine was selling for under $200 per pure gram. Importation⁷² level prices changed by less than retail-level prices. In the early 1980s, prices seemed to range between $40 and $50 per pure gram, but there were so few high-level purchases that estimates are suspect. By the late 1990s, prices seemed to be closer to $20 to $30 per pure gram.

• Figure 12 shows trends in the predicted prices per bulk gram of marijuana.⁷³ The average price per bulk gram has risen steadily from just under $5 per bulk gram in 1981 to its peak of about $15 in 1991. Prices returned close to their 1981 levels by 1998.

Indeed, price trends are broadly consistent with trends in consumption-based and supply-based estimates. During most of the 1990s, cocaine prices have been fairly constant; so too has the consumption of cocaine. During the 1990s, heroin prices have tended to fall, and relatively high-purity heroin has been increasingly available at retail. Consistent with this, heroin use appears to have increased. As noted before, marijuana use increased as marijuana prices fell, and use decreased as prices increased. Price trends are broadly consistent with consumption trends.

Figure 9—Predicted Price per Gram of Cocaine at the Retail and Importation Distribution Levels

Click for larger image

Figure 10—Predicted Price per Pure Gram of Heroin at the Retail and Importation Distribution Levels

Click for larger image

Figure 11—Predicted Price per Pure Gram of Methamphetamine at the Retail and Importation Distribution Levels

Click for larger image

Figure 12—Predicted Price per Bulk Gram of Marijuana at the Retail and Importation Distribution Levels

Click for larger image