International Journal of Drug Policy - 2014

Volume 25 Issue 3 May 2014

Issue link:

Contents of this Issue


Page 44 of 153

D. Rosenblum et al. / International Journal of Drug Policy 25 (2014) 543–555 545 segregation, in particular Puerto Rican segregation, combined with geographic proximity to Colombian trafficking routes may explain the facile entry of Colombian-sourced heroin into the eastern US in the 1990s resulting in a competitive retail market. Thus, the second part of our approach is to quantitatively estimate the relationship between patterns of ethnic segregation and: (a) the entry of Colom- bian heroin into the US and (b) the retail price per pure gram of heroin in 21 Metropolitan Statistical Areas (MSAs). 1 Methodology 1: ethnographic data The ethnographic data in Philadelphia consists of over 1300 pages of field notes and almost 2000 pages of transcribed and coded (in Atlas.ti) audio-interviews that were generated from over 430 text and audio files, 500 high-quality photographs, and 18 h of video. We collected our data from the fall of 2007 to the fall of 2012 in the classic anthropological fashion of participant-observation, requiring long-term immersion in the social world of one's research participants. To accomplish this, GK and FM lived full-time at the ethnographic field site located in the heart of Puerto Rican Philadelphia's sprawling open-air drug market. We conducted our interviews in a conversational format in the natural environment (street corners, shooting galleries, etc.) of our respondents. We pur- posefully accompanied our research respondents in their daily and nightly activities so as to minimize socially desirable responses and to maximize opportunities for direct observation of the practices and dynamics they were describing to us. Participant-observation and long-term full-time (in the case of GK and FM) immersion in street scenes enables the triangulation (over time, locations, con- texts and conversations) of self-reports by direct observation of actually occurring practices. This is especially useful among hard- to-reach or hidden populations involved in drug use and sales who are subject to pursuit by law enforcement, moral censure by the broader public and are often alienated from public health and service institutions. Methodology 2.1: quantitative analysis data description We use two sources of data to test the hypotheses derived from the ethnographic evidence. The first is the DEA's System to Retrieve Information from Drug Evidence (STRIDE). STRIDE includes data on the US heroin market from drug seizures, arrests and undercover law enforcement purchases of heroin. The STRIDE data covers the years 1990–2008. After cleaning the data using the method of Arkes, Pacula, Paddock, Caulkins, and Reuter (2004) and restricting the sample to the 21 MSAs that also have segregation data, there are 17,538 observations of retail level heroin (0.1–1 g). 2 STRIDE allows us to estimate the average price per expected pure gram of heroin at the MSA level as well as identify the place of origin of heroin obser- vations from the four major producer regions (Mexico, Southwest Asia, Southeast Asia, or South America). Almost all heroin pro- duced in South America is produced in Colombia, so we will refer to 1 This research project began purposefully as a cross-methodological multi- disciplinary dialogue to document the structural factors shaping the health risk environment of heroin injection drug users. Ethnographers (GK, FM, PB) devel- oped a theoretical explanation linking structural forces and ethnic segregation with observed data on heroin and cocaine markets in Philadelphia's Puerto Rican neighborhood. The senior author (DC), a clinician trained in both anthropological theory/methods as well as epidemiology, developed the national hypothesis and brought the anthropologists (GK, FM, PB) into analytic collaboration with a quanti- tative sociologist (JU) and economist (DR) to test the ethnography-based hypothesis quantitatively. SM provided a historian's perspective on the development of segre- gation in US cities. 2 As in Arkes et al. (2004) we restrict the sample to heroin base and heroin hydrochloride observations which make up most of the STRIDE sample. We drop observations of heroin tartrate, heroin citrate and heroin salt undetermined. Colombian-sourced rather than South American heroin throughout the paper (Paoli, Greenfield, & Reuter, 2009). From the STRIDE sample that identifies the region of origin, we take the average of the yearly proportion of heroin at the MSA- level that is identified as originating in South America and call it the Colombian Heroin Saturation Index (CHSI). The CHSI ranges from 0 to 1, with 1 indicating full market saturation and 0 meaning there has not been any observed Colombian-sourced heroin in the mar- ket. This measure is advantageous because it simultaneously takes into account the timing of Colombian-sourced heroin's entry into US cities and the extent to which it has taken over a particular market. The values in our data range from less than 0.01 (Denver, Oakland, Phoenix, and Seattle) to 0.85 (Miami). We use the estimates of the price per expected pure gram of heroin at the retail level from Rosenblum, Unick, and Ciccarone (2014), which is similar to the estimation approach of Arkes et al. (2004). The estimation process involves two steps. First, the expected purity is estimated at the MSA-year-level from the actual observed purity in the STRIDE data. This first step is necessary because buyers and possibly sellers do not know the purity of the product being purchased. We assume that buyers expect a certain level of purity when purchasing heroin which we estimate as the average observed purity in an MSA in that year controlling for the amount of heroin. 3 Second, the price per pure gram in each MSA- year at this expected level of purity is estimated for 0.5 g of heroin. 4 The second data source is derived from the US census from which we considered five measures of segregation at the MSA- level: the dissimilarity index, the isolation index, the delta index, the absolute centralization index, and the spatial proximity index (Iceland, Weinberg, & Steinmetz, 2002). Since these measures are correlated and to keep the statistical analysis as concise as pos- sible, we choose to focus on the dissimilarity index. 5 This is the most commonly used measure of segregation but more impor- tantly for our study it captures a neighborhood-level measure of segregation. The index ranges from 0 (full integration) to 1 (full segregation). When combined with the STRIDE data, we have infor- mation on both segregation and the heroin market in 21 MSAs: Atlanta, Baltimore, Boston, Chicago, Dallas, Denver, Detroit, Hous- ton, Los Angeles, Miami, New Orleans, New York, Newark, Oakland, Phoenix, Philadelphia, San Diego, San Francisco, Seattle, St. Louis and Washington, DC. The US Census Bureau provides segregation measures for American Indians/Alaska Natives, Hispanic/Latinos, Asians/Pacific Islanders, and black/African-Americans but not for subgroups within these categories. Thus we use Vargas-Ramos' (2006) esti- mates of the dissimilarity index of Puerto Ricans. The data on Puerto Rican segregation is limited in that there is only informa- tion for counties with a large population of Puerto Ricans. This limits our data to 8 MSAs where we have information on the heroin market: Boston, Chicago, Los Angeles, Miami, New York, Newark, Philadelphia and San Diego. Instead of using this data for a 3 If instead we directly measured the price per pure gram, we would have hugely inflated prices for heroin with low purity. For example, under the direct method assume a buyer gets "good" heroin half the time and "bad" heroin half the time. Paying $100 for 50% pure "good" heroin would be calculated as $200 per pure gram, while paying $100 for 1% pure "bad" heroin would be calculated as $10,000 per pure gram. On average the buyer would be paying $10,100/2 = $5050 per pure gram. Under our method, the buyer would expect purity at about 25% on average, so the price per expected pure gram would be about $400, which we believe is a much more reasonable measure of the market price adjusted for purity. Using our method consistently across MSAs allows for robust trends analysis. 4 This 0.5 g amount is simply a convenient middle point in the samples we look at which are less than 1 g. 5 The dissimilarity index is a measure of the proportion of people who would have to be redistributed within a city to achieve full integration (i.e. when there are equal proportions of different ethnicities in every neighborhood).

Articles in this issue

view archives of International Journal of Drug Policy - 2014 - Volume 25 Issue 3 May 2014