The IDA Study
Assumptions, Data, Methods, And Findings
The study by the Institute for Defense Analyses (IDA), An Empirical Examination of Counterdrug Interdiction Program Effectiveness (Crane, Rivolo, and Comfort, 1997), couples time-series data on cocaine prices and consumption in the United States with a narrative description of contemporaneous interdiction events to assess the cost-effectiveness of interdiction activities in reducing cocaine consumption. The analysis has three steps. First, a time-series index of the price of cocaine in the United States is developed. Second, this price series is associated with various time-series measures of cocaine use and juxtaposed against eight specific interdiction events. The IDA study concludes that interdiction activities generate increases in price, which in turn generates decreases in consumption. Third, the study uses its findings to examine the cost-effectiveness of interdiction activities. In particular, the cost of reducing domestic consumption by 1 percent is evaluated.
A Cocaine Price Series
The cocaine price series developed in the IDA study is based on data from the Drug Enforcement Administration's System to Retrieve Information from Drug Evidence (STRIDE). The STRIDE database contains detailed information on the quantity, price, purity, and purchase location of cocaine transactions made in undercover DEA operations since 1980.
For each transaction, a price per pure gram of cocaine is computed: that is, the ratio of the purchase price to the amount of pure cocaine received. These data on price per pure gram are then sorted by the date of purchase and divided into batches of 100 transactions each. Within each batch, the median price per pure gram is determined. This median price per pure gram is the measure used throughout the IDA study to index time-series changes in cocaine prices in the United States.
The IDA price series aggregates the prices paid for the purchase of a wide range of quantities of cocaine, from less than 1 gram to more than 30 grams. The IDA study observes that the price paid per pure gram tends to decline substantially with the quantity purchased, with the price per pure gram at the retail level (i.e., 0-10 grams) being nearly four times that at the wholesale level (i.e., 30 grams or more). However, the IDA study argues that the price series for different quantities of cocaine move proportionately over time so that when retail prices rise, there is a proportional rise in wholesale prices (Crane, Rivolo, and Comfort, 1997:II-10-13). Accordingly, the IDA study forms a single ''street-price" index that combines STRIDE transactions at all quantities (pp. II-18-19).
The Price Elasticity of Demand for Cocaine
With the price series in hand, the IDA study turns to the substantive questions of the effect of prices on consumption and the effect of interdiction activities on prices. To measure the sensitivity of demand to prices, the IDA study presents estimates of the time-series association between its street-price index and measures of cocaine use obtained from four different sources. As shown in Table 2, the four data sources measure cocaine use among emergency room patients, arrestees, workers, and patients in-treatment centers. Although these data sources provide no
TABLE 2 Estimated Price Elasticities of Demand in the IDA Study
Usage Data Set
Elasticity of Demand
DAWN (Drug Abuse Warning Network)
Emergency room admittances
Arrestees in 23 cities
SBCL (SmithKline Beecham Clinical Laboratories)
TEDS (Treatment Episode Data Set)
Patients in cocaine treatment centers
SOURCE: Estimates reported in Crane, Rivolo, and Comfort (1997:Table III-1).
information on quantities of cocaine consumed and although they concern only particular subpopulations of potential users, the IDA study argues that they provide measures of cocaine use that are "logically linked" (p. III-2) to consumption and thus provide insight into the price elasticity of demand.
The IDA study uses its street-price index and the four data sets to estimate how the number of users varies with street prices, and it interprets the findings as estimates of the price elasticity of the demand for cocaine. The four estimates displayed in Table 2 range from -0.51 to -0.71. Given these estimates, the study concludes that the conventional wisdom value of -0.5 constitutes a "reasonable estimate of the price elasticity of demand" (p. III-9). Thus, a 1 percent increase in the price is assumed to result in a 0.5 percent decline in the quantity demanded.
Effect of Interdiction Events on the Price of Cocaine
The centerpiece of the IDA empirical analysis is its juxtaposition of the street-price time series against the onset of eight major drug interdiction events identified by the historian of the U.S. Army Southern Command. Figure 4, which replicates Figure IV-1 of the IDA study, displays the price series and information on the initiation of these interdiction events. In this figure, each circle indicates the median value of 100 consecutive individual transactions; the solid line is derived from a nonparametric fit to these data. The data show that the basic downward trend in cocaine prices is interrupted by an abrupt and lasting change in 1989 and by a number of short-lived upward "excursions" that appear from time to time. The IDA study attributes these interruptions to the eight interdiction events, stating "the sudden change in the price decay rate and each of the short-term excursions are shown to follow the initiation of major interdiction activities, primarily in the source zone nations, and are thus to be causally connected" (pp. 1-2). For example, the sudden and lasting change in the price decay rate that occurred in 1989 is attributed by the IDA study to the interdiction activities and infrastructure developed by the Bush Administration's "war on drugs."
Cost-Effectiveness of Interdiction
The IDA study uses the structural change to the cocaine market that it identifies as having occurred in 1989 to examine the overall cost-effectiveness of interdiction activities. Figure 5, which replicates Figure IV-5 of the IDA report, displays an extrapolated trend curve for the street price of cocaine. This trend curve is computed under the assumption that price follows an exponential path over time. Note that the line fits the data well
from 1983 to 1989, but that actual street prices lie above the trend line after 1989. The IDA study asserts that in the absence of interdiction activities, especially the war on drugs, street prices would have continued to follow the exponential path after 1989. Therefore, the argument continues, the difference between the actual street price series and the trend curve "is an overall measure of the efficacy of the entire interdiction effort in the transit and source zones" (p. IV-15). By this measure, interdiction activities raised street prices at least 60 percent, from about $35 per gram to $55 per gram.
The final step in the IDA chain of reasoning combines the study's finding on the price effect of interdiction events with the conventional price elasticity of demand estimate of -0.5 and with data indicating that the U.S. government spends about $1-2 billion per year on all interdiction activities. According to the IDA report, interdiction activities increase street prices by 60 percent, which in turn results in a 30 percent reduction in the quantity of cocaine demanded. The study concludes that cocaine consumption falls by about 0.015 percent for every $1 million spent on interdiction activities. Hence, to reduce cocaine consumption by 1 percent, spending on interdiction would need to increase by $30-60 million (Crane, Rivolo, and Comfort, 1997:IV-15).
The IDA study provides a wealth of information on the time series of cocaine prices, purity, and use in the United States since 1980. The study displays a set of detailed graphs that provide an informative summary of the primary, available time-series statistics. The study uses these statistics to conjecture an intriguing association between eight major interdiction events and short-term fluctuations in cocaine prices.
These positive features notwithstanding, the IDA study makes many questionable inferences about the effects of interdiction on the cocaine market in the United States. The primary shortcomings of the study are the conclusions drawn using the statistics presented. There are many plausible alternative interpretations of the price fluctuations found by the IDA study. It may be that the short-term price fluctuations in the IDA street-price series are artifacts of the STRIDE data acquisition procedures and are unrelated to substantive characteristics of the market for cocaine. It may be that the apparent long-term interruption to the price trend in 1989 was due to the destruction of the Medellin cartel by the Colombian authorities, rather than to the interdiction activities carried out during the war on drugs. With the exception of the 1989 interruption, none of the price changes that the IDA study attributes to interdiction events was
large, and none of the price changes lasted longer than approximately 1 year. Thus, even if one accepts IDA's interpretation of the data, the effects of interdiction are small and temporary.
The committee's major concerns about the data used in the IDA study include the development of the price series and the procedure for selecting interdiction events. Even if one accepts the IDA data at face value, the committee has deep reservations about the inferential methods used to relate price fluctuations to intervention events. And, even if one accepts the IDA data and inferential methods, the committee notes that the study at most shows that interdiction events have temporary effects on cocaine prices. Each of these concerns is detailed in the rest of this chapter.
The committee concludes that the main contribution of the IDA study is to provide descriptive time-series statistics on the domestic cocaine market. The study's findings on cost-effectiveness are not persuasive because they are based on data and inferential methods of questionable validity. One cannot reasonably infer the cause of an event from time-series data without accounting for other factors that might influence that event. It may be that the observed relationship between price fluctuations and interdiction events is causal, but the IDA study does not support this conclusion.
The IDA Price Series
The STRIDE price data are essential to the IDA analysis. The IDA findings on the effectiveness of interdiction rest critically on the proximity in time between interdiction events and upward excursions in the IDA street-price series. It is not clear whether the purported excursions are real or are artifacts of the STRIDE sampling process, of the IDA procedures transforming the STRIDE data into a price series, or even of random sampling error.
The STRIDE Data
Beginning in 1971, the Drug Enforcement Administration (DEA) has used the STRIDE data to provide information on seizures and undercover purchases of illegal drugs made in the course of DEA operations. The STRIDE data give the locations and dates of seizures and undercover purchases and measure the quantity and purity of the drug.
The STRIDE data on undercover purchases of cocaine are potentially useful to the study of the cocaine market because they reflect actual market transactions. However, these data pertain to transactions of many different quantities of cocaine of varying purity. Hence, the data obtained
from different transactions must somehow be standardized if they are to be comparable (Office of National Drug Control Policy, 1998; Caulkins, 1995b; Frank, 1987).
A vexing problem for analysis of the STRIDE data is that the observed transactions do not constitute a random sample of cocaine transactions. The sampling process that generates the STRIDE data is not well understood. The IDA study asserts that the nonrandomness of the STRIDE sample of cocaine prices and the resulting concern about the unrepresentative nature of the sample do not matter if the STRIDE sampling process is "approximately stationary in time" (p. II-3). The study does not adequately explain what it means by this assertion, provide evidence that the assertion is valid, or explain how the validity of the assertion would eliminate concerns about the nonrandom nature of the sampling process.
Among analysts of the cocaine market, there seems to be agreement that the STRIDE data satisfactorily describe long-term trends in cocaine prices, such as the drop by approximately a factor of six in the period 1983-1990 (Riley, 1996; Caulkins, 1994, 1995b; DiNardo, 1993). Most of the IDA analysis, however, is concerned with short-term price fluctuations, not with long-run trends. Even if the STRIDE data describe long-term trends satisfactorily, they may not accurately describe short-term fluctuations. It may be, for example, that short-term variations in the geographical distribution or quantity distribution of the cocaine purchases made by undercover agents produce short-term fluctuations in STRIDE prices of the magnitudes reported by IDA for the 1992-1997 period. In the absence of information on the DEA's purchasing strategy, it is impossible to know the extent to which the short-term price fluctuations identified by IDA are artifacts of changes in undercover purchase patterns and the extent to which they describe real changes in cocaine prices.
The Street-Price Index
Even if the STRIDE data did describe a random sample of cocaine transactions, the street-price index developed in the IDA study would be highly suspect. The IDA street-price index is the median price per pure gram for batches of 100 DEA purchases that occur at about the same time. The IDA argument that prices across different purchase quantities may be aggregated rests largely on the apparent proportional movements in prices for different quantities that occurred in the period of the war on drugs. Visual scrutiny of the graphs displayed in the IDA study does not, however, indicate clear co-movements of this type during other periods. Moreover, the study does not carry out any formal statistical procedures
to evaluate the degree of co-movement among the price series for different purchase quantities.1
Even if the IDA study is correct in claiming that prices at different quantities move proportionally over time, the aggregation of prices per pure gram into a street-price index remains problematic. Variation over time in the quantity-distribution of nationwide cocaine purchases can yield variations over time in the IDA street-price index even if the price per gram at each purchase quantity is constant over time. The IDA study asserts that "care was taken to ensure that no major changes in the purchase volume distribution have occurred over the period of interest" (p. II-19), but no details are given. The study does not specify what is meant by "major changes" nor what "care was taken."
The IDA study aggregates cocaine prices to reduce the statistical variability in the price series (II-18, 19). However, the procedure used in the IDA study to construct a street-price index makes it difficult to assess the validity of the study's findings on the association between interdiction events and short-term price fluctuations. It would have been better to construct a weighted price index in which the weights reflect variations over time in the nationwide quantity distribution of purchase quantities and variations over time in the DEA sampling process. In the absence of more convincing evidence about the co-movements of the price series across different quantities, an even better alternative would have been to separately analyze price series for different quantities and geographical strata.2
It should be noted that the IDA study's procedure for constructing a price index for cocaine does not conform to standard approaches. Standards for constructing retail and producer price indexes have been developed by scholars and by the Bureau of Labor Statistics of the U.S. Department of Labor over many years. The standards are incorporated in such measures as the consumer price index (CPI) and the producer price index (PPI). The IDA study does not draw on this body of knowledge.
Purity as a Measure of Product Quality
A common problem in the construction of price indices is how to appropriately account for variations in product quality. The IDA study identifies cocaine quality with DEA laboratory measurements of purity and measures price as price per pure gram.
This procedure is sensible if one assumes that the participants in market transactions of cocaine have full knowledge of the purity of the product they are buying or selling. This assumption may be reasonable with respect to wholesale transactions of cocaine, but it is questionable with respect to retail transactions. It is especially questionable whether the consumers of cocaine, who may engage in many small transactions with different sellers and who do not ordinarily have access to laboratory equipment, are perfectly aware of the purity of the product they are purchasing in any given transaction.
If consumers do not know the purity of the product they are purchasing in given transactions, the demand for cocaine need not vary with price per pure gram in the simple downward sloping manner of classical demand analysis. Consumer behavior when presented with seller's offers of cocaine would depend on what the consumers know about purity. Do they know the distribution of purity of the cocaine offered by a particular seller? Do they only know the distribution of purity of all the cocaine that comes to market? Do they misperceive the distribution of cocaine purity? If the distribution of purity varies over time, what is the process by which consumers learn about these variations? Answering these questions is important to the analysis of the demand for cocaine and to the construction of appropriate price indices.
Even if the STRIDE data did provide a random sample of transactions and an appropriate price series were formed from these data, a statistically valid analysis of the price series would require due attention to sampling error. The IDA study reports that cocaine prices are highly variable across transactions even within quantity strata (p. II-18). Nevertheless, the study provides no statistical measures, such as standard errors or confidence intervals, to describe the precision with which its street-price series is estimated. In the absence of such measures, it is impossible to interpret the price fluctuations that appear in Figure 4. Do these fluctuations reflect real variations in cocaine prices or are they simply the consequence of sampling error? The IDA study provides no information that would enable a reader to answer this most basic question.
Selection of Interdiction Events
A serious difficulty in interpreting the IDA study is the dearth of information provided on the eight "major" interdiction events whose occurrence is juxtaposed against the time series of street prices. The study does not describe the criteria used to select these events. Without this information, interpreting the relationship between interdiction activities and street prices is problematic. The only information revealed in the IDA study is that the historian of the U.S. Army Southern Command identified 10 major actions. Two of these actions were not included in the analysis because "further examination has shown that these operations were not, in themselves, major operations in the source zone" (p. IV-4). What does it mean to be a "major" action? Are actions deemed to be ''major" ex ante, or are they so deemed ex post if they are observed to be sufficiently successful?
In an effort to obtain answers to these questions, the committee requested data from and a briefing by IDA personnel on any and all information pertaining to the selection of interdiction events. The classified briefing was held in December 1998: present were the IDA authors of the study and other IDA personnel; officials from the U.S. Department of Defense, the Drug Enforcement Administration, the White House Office on Drug Control Policy, and the Central Intelligence Agency; and members of the committee's classified subcommittee. However, neither the written materials nor the oral briefing provided answers to the committee's questions: what criteria were used for classifying events as "major" and whether such classification was made prior to or after an event. As noted above, one cannot interpret the relationship between interdiction activities and the time series of street prices without that information.
Whatever the IDA criteria may be, the committee's central concern remains that the IDA study only examines the possible effects of eight specific interdiction events and not of interdiction activities in general. At worst, if interdictions are classified as major events on the basis of ex-post realizations of street prices, the observed associations are meaningless. At best, focusing on only eight events implies that the IDA study reveals the effects of the eight events identified. The study does not offer evidence on the overall effects of interdiction activities on the cocaine market. The effects of the many "nonmajor" actions that consume most of the interdiction budget remain unknown.
The IDA study uses its cocaine street-price series coupled with time-series information on domestic cocaine use and major interdiction events
to draw inferences about the responsiveness of demand to changes in prices and about the responsiveness of prices to interdiction activities. The inferential methods used are appropriate only under extremely rigid assumptions about the behavior of cocaine demand, supply, and prices over time.
The IDA analysis attributing price changes to interdiction events and evaluating the cost-effectiveness of interdiction in reducing cocaine consumption implicitly assumes that the demand curve relating quantity of cocaine to price was fixed during the entire 1983-1996 period examined. That is, the study assumes that the quantity of cocaine consumed in the United States would have remained constant during this period if no changes in cocaine prices had occurred.3 The analysis also implicitly assumes that supply-control activities other than interdiction remain constant over time. Going even further, the IDA study assumes that the time series of cocaine prices would follow an exponential path if it were not for the mediating influence of intervention events.
None of these extremely strong assumptions is substantiated in the study, and none of them has a priori plausibility. The failure of any one of them might be enough to overturn the study's major findings.
Attribution of Price Fluctuations to Intervention Events
The IDA study asserts without substantiation that all deviations in price from the exponential decay path shown in Figure 5 should be attributed to interdiction events. The study maintains that there are no other plausible causes of observed price fluctuations (Crane, Rivolo, and Comfort, 1997:IV-5, IV-10).
One obvious potential source of price fluctuations is time-series variation in the demand for cocaine. It seems likely that the demand for cocaine has changed over time. During the period of the IDA study, attitudes toward cocaine consumption changed, the mix of light and heavy users changed, and the penalties for cocaine possession changed. Any of these factors would probably shift the demand for cocaine and, hence, might have caused price fluctuations.
Another obvious source of price fluctuations is variation in the supply of cocaine for reasons other than interdiction activities. The IDA
study attributes the large price increase in 1990 to the war on drugs, but as noted above it may have resulted from the dismantling of the Medellin cartel by the Colombian government. In addition, the study attributes the price excursion observed in 1995 to the initiation of an aircraft shoot-down policy in Peru, but the timing makes it likely that other factors were at work as well. Figure IV-2 of the IDA study shows that the quantity of coca base being transported across the air bridge from Peru to Colombia had been decreasing steadily for more than 2 years before the Peruvian government began shooting down aircraft, and the decrease accelerated sharply approximately 6 months before the shoot-down policy went into effect. The price of Peruvian coca base also decreased sharply 6 months before initiation of shoot-downs (Figure IV-3). The timing of these events makes it implausible that the shoot-down policy is the sole factor for the price excursion identified as occurring in the United States in 1995.
Time-series variations in domestic enforcement activities may also have generated price fluctuations. Many important changes in domestic enforcement policy occurred during the same periods as the eight interdiction events that are the focus of the IDA study. The Crime Control Act of 1984 established tougher penalties for drug-related offenses. The 1986 and 1988 Anti-Drug Abuse Acts increased domestic law enforcement efforts as well as demand reduction programs. In the latter part of the 1980s, the number of people sentenced to prison for drug-related offenses increased substantially (see Beck and Gilliard, 1995; Beck and Blumstein, 1999; U.S. Sentencing Commission, 1998).
The IDA study makes essentially no attempt to substantiate its assertion that interdiction events have been the only source of fluctuations in the price of cocaine in the United States. The price-series data alone simply cannot reveal whether the prices fluctuations that the study attributes to the eight interdiction events should so be attributed or were instead due to other contemporaneous changes in market supply or demand.
The Exponential Price Path
The analysis that attributes price fluctuations to interdiction and the subsequent analysis of the cost-effectiveness of interdiction in reducing cocaine consumption rely on acceptance of the exponential price path displayed in Figure 5. By construction, the fitted exponential path closely tracks the decline in cocaine prices in the 1980s. The extrapolated path lies below the realized price series after the abrupt change that occurs in 1989. In fact, it is not known why cocaine prices decreased sharply before 1989 nor why they became relatively stable in the 1990s. There is no reason to believe that the natural path of prices is exponential. The expo-
nential path and its extrapolation after 1989 are speculative and, in the absence of corroborating evidence, should be disregarded. In particular, there is simply no basis for assuming that in the absence of interdiction activities the street price of cocaine would have fallen to $35 per pure gram, as the IDA study asserts.
Duration of Price Excursions
The price excursions evaluated in the IDA study appear to be small in magnitude and short lived. None of the price changes that the IDA study attributes to interdiction events come close to the magnitude of the unexplained six-fold drop in price that took place between 1983 and 1990, presumably despite vigorous enforcement action. Furthermore, the price excursions identified in the IDA study are all temporary. The study argues that some of the interdiction events had short durations, but others were sustained efforts to which the cocaine production process eventually adjusted. That is, the price increased for some time but eventually decreased again.
Consider, for example, the implementation in 1995 of the shoot-down policy over the Peruvian-Colombia air bridge. Figure IV-2 of the IDA study indicates that shipments on the air bridge did not recover by the end of the observation period in mid-1996, but the street price of cocaine did return to its pre-shoot-down level. According to the IDA study, the production process may have adapted after about a year; alternatively, the price may have fallen in 1996 because of the takedown of the Cali cartel leaders (p. IV-3).
The cost-effectiveness estimates reported in the IDA study do not account for possible dynamic relationships between prices and events. Cocaine producers whose shipments are interrupted by interdiction activities have an incentive to develop alternative supply sources and shipment routes. The cost-effectiveness estimates in the IDA study implicitly assume that no good alternatives are available, so that the supply curve is close to vertical. This may perhaps be a reasonable short-run assumption, but it is highly questionable as a long-run assumption.
The IDA study is best thought of as a descriptive time-series analysis of statistics relevant to analysis of the market for cocaine in the United States. The study makes a useful contribution with a wealth of empirical, time-series evidence on cocaine prices, purity, and use since 1980. Efforts to understand the operation of the market for cocaine must be cognizant of the empirical data, and the IDA study presents many of these data and
calls attention to some intriguing empirical associations among the various series.
However, the IDA study does not yield useful empirical findings on the cost-effectiveness of interdiction policies to reduce cocaine consumption. Major flaws in the assumptions, data, and methods of the study make it impossible to accept the IDA findings as a basis for the assessment of interdiction policies. For example, the conclusions drawn from the data rest on the assumption that all time-series deviations in cocaine price from an exponential decay path should be attributed to interdiction events, not to other forces acting on the market for cocaine. Numerous problems diminish the credibility of the cocaine price series developed in the study, and an absence of information prevents assessment of the procedure for selecting interdiction events.