I. An Analytic Framework

The principal costs of the international drug industry (1) are associated with distribution rather than production; Table 1 provides fairly standard figures on the cost of cocaine at different points in the distribution system and generates three observations, which are also true for heroin:

1. The cost of production, as opposed to distribution, is a trivial share of the final price. That statement holds true even if one includes the cost of refining as well as leaf production.

2. The vast majority of costs are accounted for by domestic distribution in the United States.

3. Most of the domestic distribution revenues go to the lowest level of the distribution system. If the retailer and lowest level wholesaler each raise their purchase price by 75 per cent, they will account for 60 per cent of the final price.

Risks and the costs of bearing them provide a reasonable explanation for these observations. Coca and opium are grown in countries characteracterized by labor and land that have low prices relative to those in Europe and North America. The comparative advantage of these countries is reinforced by the reluctance of governments in Bolivia and Peru (for coca) and Burma, Afghanistan and Laos (for opium/heroin) to act aggressively against growers or early stage refiners. Low opportunity cost for factors of production plus low enforcement risks produce very modest prices for the refined product and also ensures that production does not move upstream geographically.

It is also useful to consider why neighboring countries, involved in trafficking, are not major producers. Both Colombia and Thailand have become relatively rich, raising the opportunity cost of land and labor, thus, they may not be able to compete in the leaf or opium growing sector given that the illegality of the product has inhibited the development of more technologically advanced growing methods. Both Colombia and Thailand, despite the corruption of their drug control efforts, have also been more willing to act aggressively against growers; Colombia indeed allows spraying of coca fields.

The modest share of costs associated with cocaine smuggling is easily explained. (2) The drug travels in large bundles at that stage; seizures suggest that shipments of 250-500 kilograms are quite common. Though large sums may be paid to pilots for flying small planes carrying cocaine or for Honduran colonels for ignoring their landings, these costs are defrayed over a large quantity. A pilot who demands $ 500,000 for flying a plane with 250 kilograms is generating costs of only $ 2,000 per kilogram, less than 2 per cent of the retail price. Even if the plane has to be abandoned after one flight, adds only $ 2,000 to the kilogram price.

Heroin smuggling is not so efficient. Because most of the low cost producers are very distant from the United States. dedicated vessels cannot be used, as they are in the cocaine trade. A great deal of heroin travels in small bundles carried on invidividual couriers. (3) There have been a few multihundred kilogram shipments of heroin but they are very rare compared to those for cocaine; (4) correspondingly, the mark-up between export and import prices for heroin is much larger. Nonetheless, body-packing, where the body-packers are low wage earners, produces per kilogram smuggling costs of less than $ 10,000. A body-packer can apparently carry about 3/4 of a kilogram. A payment of $ 5,000 for incurring a 1 in 10 risk in prison (perhaps acceptable for couriers whose legitimate wages are only about $ 2,000 per annum), along with $ 3,000 in travel expenses, produces a kilogram cost of just over $ 11,000 (5) compared to a retail price of $ 500,000.

Mexico of course represents a different situation with respect to smuggling costs; the land border allows for much cheaper importation. However, Mexico is a high cost producer; farm gate prices for opium in Mexico are typically $ 2,000 to $ 5,000 per kilo, compared to a few hundred dollars in Burma; hence the low smuggling costs simply equalize total costs. Colombian source heroin, a new source, also represents high farm gate production with relatively low smuggling costs.

Finally, the high costs associated with the final distribution activities point to where the high risks are. A 1990 study estimated that the annual risk of imprisonment for a low level drug dealer in Washington, DC was about 2 in 9; the expected time served was 18 months. (6) In addition, low level dealers then faced a 1 in 70 annual chance of beign killed and a 1 in 14 risk of serious injury. These risks are defrayed over small quantities of drugs, since each retailer sells only a few grams per week. If a 1/4 gram (pure) cocaine sale exposes a seller to a 1 in 3,000 risk of going to prison and a 1 in 70,000 probability of getting killed (consistent with the findings of the study), then a mark- up of $ 10 is not unreasonable. At $ 40 per gram, this explains the high share of the total funds going to dealers at the end of the distribution system. The great fortunes may be made at the high ends of the trade but the sharp pyramiding of numbers, with a hundred gram dealers for every kilogram dealer, means that most of the money goes to moderate incomes for low level dealers.

Implications for International Drug Policy

Simple and sketchy though this account is, it does have some implications for explaining the location and organization of the drug trade. Smuggling costs depend on the ability to conceal drugs in a flow of legitimate commerce and traffic. Let us consider Nigeria's role.

Nigeria is a nation that seems to have little potential role in the international drug trade. It is isolated from any of the principal producer or consumer countries and lacks a significant base of traditional domestic production or consumption. Nonetheless, Nigerian traffickers have come to play a significant role in the shipping of heroin between Southeast Asia and the US; recently these traffickers have even entered the cocaine business, which is still more remote from their home country.

The explanation is to be found in a complex of factors. Nigerians are highly entrepreneurial, have been misruled by corrupt governments over a long time, have large overseas populations, very low domestic wages (by international standards) and moderately good commercial links to the rest of the world. Thus it is relatively easy to buy protection for transactions in Nigerian airports (corruption and a weak governmental tradition), to establish connections in both the source and consumption nations (large overseas populations) and to use.

One might more usefully ask whether the new republics of Central Asia are likely to become major players in the international heroin business. They certainly have low cost land and labor, as well as apparently good ecological conditions for growing opium and a traditional experise. Some governments, such as Uzbekistan, are desperate for foreign currency, have few alternative sources and little concern about their standing in international organizations; they are unlikely to aggressively enforce prohibitions against growing opium poppies or to have the capability to do so even if they desired to. They are certain to be low cost producers.

But they are advantaged, compared to current low cost producers, notably Afghanistan and Burma? Though closer to Europe and perhaps with significant populations resident in Russia and perhaps even in Western Europe, the commercial connections with Western Europe are likely to be weak compared to Burma,(through established Thai and Chinese trafficking networks), imbedded in growing legitimate traffic. Thus I suggest that the Central Asia republics will only become major players in the European opium markets if there are disruptions (including rapid economic development) in the current major supplier countries.

II. Measurement The US devotes considerable resources to estimation of drug production by nations. (7) Those estimates, published each year in the 'International Narcotics Control Strategy Report' ('INCSR'), are essentially without competition; certainly they are regards as more authoritative than any other reports, such as the documents of the United Nations Drug Control Program. And some notion of the scale and rate of change of the drug trade is potentially important for decision makers. However, the current system produces numbers that have unnecessarily low credibility and are detached from the policy process they are supposed to inform. The low quality shows particularly in inconsistency over time and across sectors of the industry; the policy irrelevance shows in the lack of policy responsiveness to changes in the estimates.

Consider for example estimates of Burmese opium production, consumption and export. In the 1991 'INCSR' the opium available for refining (primary into heroin) for 1988 was estimated to be 679 metric tons; the figure for 1989 was 1,600 metric tons. The difference reflected a 25 per cent increase in cultivation, an unexplained increase in yield per acre and a dramatic reduction in the amount of opium exported; heroin production was reported to thus increase from 68 tons to 128 tons. Then in the 1992 'INCSR' the figure for domestic consumption in 1988 was suddenly reduced from 400 tons to 150 tons, reflecting a downward revision in the number of Burmese opium users from 400,000 to 34,000. Yet this did not lead to any change in the estimate of the amount available for refining in 1988. A 25 % increase in the estimated number of Burmese heroin users from 12,000 in 1988 to 15,000 in 1989 led to a more than doubling of the estimate in Burmese domestic heroin consumption from 2.0 tons to 4.5 tons. (8) These are figures that do not bear close scrutiny, either individually or collectively.

Mexican Marijuana (9)

The problems are illustrated even more graphically by the published estimates for Mexican marijuana production in the late 1980s. These estimates have relied on often inconsistent and inadequately described methodologies, leading some analysts, including one at INM, to conclude that actual production is ''unknowwable'' and that the agency's estimates are at best rather unscientific guesswork. (10) The preface to the 1988 annual report of the National Narcotics Intelligence Consumers Committee (NNICC, an interagency group chaired by DEA) warns that ''there is little reliable data upon which to base estimates of the quantities of drugs involved.'' (11) Yet, notwithstanding the lack of first hand evidence of illicit activity that limits accurate production measures, it is often claimed that ''the general trends portrayed can be considered reliable.'' (12) This was belied by discrepancies in the trends for the two official series for these estimates, provided by NNICC and the 'International Narcotics Control Strategy Report.' For example, the NNICC estimate of net production (after eradication) rose in 1986 to 5,460 metric tons from 4,125 the year before; the 'INCSR' estimates for the same years were 2,800 in 1986 and 2,700 in 1985. Note that they differ substantially both in absolute value and in the trend; the NNICC figures are higher and rising while the 'INCSR' figures are lower and essentially flat. In 1987 INM showed a slight decrease (about 5 per cent) while the NNICC showed an increase of about one third. INM officials explained the discrepancy in the mid-1980s as follows: ''The Department of State considers its country estimates more reliable because the data were derived principally from aerial surveys. There are, however, no survey data on marijuana cultivation in Mexico; the State Department relied on random reports from Mexico which were higher than the NNICC figure, which is an extrapolation of seizure data.'' (13)

The fundamental unsoundness of the whole series of estimates was demonstrated vividly at the end of the decade. The US estimate of Mexican marijuana production was dramatically increased in the 1990 'INCSR,`from a total of 5,700 tons in 1988 to 47,000 tons in 1989, as the result of changes in estimation techniques. (14) No details of those changes were provided in the published document. Yet it was possible to determine, with no great technical skill, that these figures were implausibly high and should never have been published.

Consider the various ways that Mexican marijuana might have been disposed of: seizues, domestic Mexican consumption, exports to Europe and exports to the United States. Mexican domestic consumption is thought to be quite low, notwithstanding the association of marijuana in the US with Mexican immigrant groups in the 1930s. The State Department estimated the total in the late 1980s to be 100 tons. There are no reports of Mexican exports to European markets, probably because it is not well located to compete with North African, Middle Eastern and domestic production in Europe, which is primarily a hashish rather than marijuana market anyway. Seizures are usually estimated at a few hundred tons. Nonetheless, assume that all these figures are major underestimates and that the total for seizures, Mexican consumption and European exports was 12,000 tons.

This would leave 35,000 tons for consumption in the United States markets. How many marijuana users would have to purchase the Mexican product to dispose of this? Using data from the annual high school senior survey in the early 1980s, I estimated that the average joint contained about 0.4 grams of marijuana. (15) By the late 1980s, average marijuana potency had risen substantially, (16) probably reducing the amount of marijuana used in each joint. However, let us assume that there was no such increase for Mexican origin marijuana and that Mexican marijuana is of lower potency, so that each joint contains 1 gram.

A very heavy user of marijuana consumes about 3 joints per day. Giving him time off for colds and work related drug tests, he consumes this amount 333 days a year; this (conveniently) gives a total of 1 kilogram of marijuana annually. That implies that we need 35 million very heavy marijuana smokers to consume 35,000 tons; that would be about half of all persons aged 12-35, the heavy user ages. And this does not take into account consumption of domestic US production or what is imported from Jamaica, Colombia etc. Yet reasonable estimates of the total number of heavy users (at least one joint per day) are only about one tenth of the 35 million needed to dispose of the imports from Mexico. Moreover, other estimates of US consumption have been far lower; Kleiman came up with figures for 1985 of only about 5,000 tons (17) and prevalence was still declining in the late 1980s.

The 1991 'INCSR' announced a further revision in estimation methodology. The estimate of area harvested was increased, reflecting a dramatic downward revision in the estimation of acreage eradicated. However, a new distinction was introduced between ''usable plant yield`` and ''whole plant yield``; the former, more relevant to consumption estimates, was put at only half the latter. The new 1989 estimate of usable plant available for export after domestic consumption and Mexican seizures was 29,700 tons. Though an improvement over the previous figure, it was still utterly implausible, requiring US consumers to account for far more than is consistent with current estimates of prevalence. Only in 1991 did the official figure start to approach plausible levels.

Though I will not describe all the twists and turns since then, it is worth noting the current state of estimates. The 1994 'INCSR' lists the following series for usable plant yield:

		1989		1990		1991		1992		1993
metric tonnes		30,200		19,700		7,795		7,775		6,280

Only for the 1990 figure is any footnote explanation offered suggesting a change in methodology; yet no one seriously maintains that 1993 Mexican cannabis production is 20 per cent of its 1989 level. The series should at least have been made consistent.

I suggest that these figures should never have survived a review process to the point of publication. As the above analysis suggests, it is easy to establish that they are far outside the plausible range. Yet the estimation process is so detached from analysis of domestic indicators that these figures have been able to survive for many years. Colleagues of mine at RAND have recently developed elementary ''mass conservation'' models that impose consistency checks on estimates; (18) the discipline is simple enough that one may reasonably ask why it has not been done before.

The story also points to another aspect of detachment, important for those interested in estimates of foreign production. The increases and declines in the 'INCSR' estimates have no clear consequence for US policy. Congress did not feel the need to take new measures against Mexico when suddenly the State Department produced figures suggesting that the US marijuana market was completely dominated by Mexico; nor did the administration make any noticeable change in policy. The current drug production estimates are detached from the policy process; one reason they are of low quality is that they simply have no consequence for any senior decision makers.

The State Department is required by statute to produce these figures annually. It cannot simply claim an incapacity to meet the statutory requirement. Nor, I suspect, is there any justification for large investments in improving the available data. There is no excuse however for producing implausible figures or inconsistent series.

What is needed by way of measurement? Clearly this is entirely a policy driven enterprise, without any scientific goals, as is not the case for the collection of data on domestic use and abuse. So the central question, putting aside the detail of the statutory requirements imposed on the executive branch by the boundlessly ambitious Congress, is what decisions these figures inform. At the moment only the certification process seems to be connected to these estimates and that is actually driven by politics rather than numbers; i.e. the administration forgives

US friends and condemns those with whom it is not on goods terms (e.g. Burma, Iran and Syria). Without a specification of policy needs, the estimates will continue to flounder.

Notes

(1) This paper confines itself to cocaine, heroin/opium and marijuana; other illicit drugs, primarily synthetics, are sources of significant problems but figure much less prominently in international trafficking as far as the United States is concerned.

(2) This analysis draws heavily on Reuter ''Can the borders be sealed?'' 'The Public Interest' 1988.

(3) Nigerian traffickers seem to specialize in such smuggling. Mark Kleinman has estimated that Nigerian couriers body packing heroin into New York account for over 500 kilograms per annum.

(4) These large shipments may account for a similar share of total US consumption of heroin, which is only about 10 per cent of cocaine consumption. However the shape of the shipment size distribution seems to be different.

(5) The risk and payment figures here are moderately informed guesses; the purpose is simply to provide a sense of the magnitudes involved.

(6) Reuter, P.; MacCoun, R. and P. Murphy 'Money from Crime: The Economics of Drug Selling in Washington DC' Santa Monica, Ca., RAND, 1990.

(7) I infer this from the publications and contacts over the years with participants in the estimation process. No cost figures have ever, to my knowledge, been published.

(8) The implied annual heroin consumption per addict is 300 grams, 10 to 20 times the figure for the United States. No doubt heroin is vastly cheaper in Burma but incomes are also dramatically lower.

(9) This section draws extensively on Reuter, P. and D. Ronfeldt 'Quest for Integrity: The Mexican-US Drug Issue in the 1980s' Santa Monica, Ca., RAND, 1992.

(10) See Kleiman, M. 'Marijuana: Costs of Abuse, Costs of Control' Boulder, Colo., Westwood Press, 1989; personal communication from an INM official, 1989.

(11) Drug Enforcement Administration 'National Narcotics Intelligence Consumers Committee Report' 1988.

(12) Drug Enforcement Administration 'National Narcotics Intelligence Consumers Committee Report' 1984.

(13) General Accounting Office 'Control: Drug Interdiction and Related Activities along the Southwestern US Border' Fact Sheet GAO/GGD-88- 124FS Washington, DC September 1988, p. 53.

(14) ''New analytic methodologies have enabled the US government to assess more accurately the extent of marijuana cultivation during the past several years'' ('INCSR,' 1990, p. 13). The report included no revision of previous years estimates.

(15) Reuter, P. ''La signification economiques des marches illegaux aus Etats-Unis: le cas de marijuana'' om Archmabualt, E. and X. Greffe (eds.), 'Les Economies Non Officielles,' Paris, Editions Le Decouverte, 1983.

(16) Moore, M. ''Supply reduction and drug law enforcement'' in Tonry, M. and J. Wilson (eds.), 'Drugs and Crime,' Chicago, University of Chicago Press, 1990.

(17) Kleinman, M. 'Marijuana: Costs of Abuse, Costs of Control' Boulder, Colo., Westwood Press, 1989.

(18) Separate models are available for cocaine, heroin and marijuana. Resetar, S. and B. Dombey-Moore 'A System Description of the Cocaine Trade' Santa Monica, Ca., RAND, 1994. Childress, M. 'A System Description of the Heroin Trade' Santa Monica, Ca. 1994. Childress, M. 'A System Description of the Marijuana Trade' Santa Monica, Ca. 1994.