Discussion
Drug Treatment Data
Health Data: Invisible Populations
One major problem with our analysis is the incomplete nature of the attribute data. As stated in the analysis, SPA health care data was drawn from the Los Angeles County Department of Health Services‘ Alcohol and Drug Program Administration (ADPA) annual report for the 2002-03 fiscal year. It reviewed the 122 Alcohol and Drug (AOD) programs contracted by the LA County ADPA to provide services to 45,747 participants (APDA, 2003, p. i).
Participants came from a wide range of backgrounds. Nine major networks of service provision comprise the AOD Program, and they cater to youth, welfare reform support services recipients, non-violent drug defendants, male and female inmates paroled to Los Angeles County, welfare recipients, opiate-dependent individuals undergoing detoxification or maintenance, pregnant and parenting women, and Proposition 36 offenders (adult criminal defendants convicted of non-violent drug offenses) (APDA, 2003, p. 4)
Participants are those who are identified by either the Los Angeles criminal or social services systems as having an alcohol or drug-related problem that needs treatment. Therefore, not every person with an alcohol or drug problem in Los Angeles county is represented by this data. The method by which the data was collected disproportionately represents those at the lower levels of the socioeconomic strata because they lack personal resources and thus need the services, such as welfare, welfare reform or pregnancy assistance, that identify their drug problem and require them to attend treatment programs. Statistics used in this analysis represent some people whose drug problems were identified not because they were arrested for buying drugs or in a drug-related crime, but because they needed social assistance and were then found to need drug-addiction treatment as well.
Los Angeles uses its criminal and social services systems as a funnel through which they direct their clients to treatment facilities. Individuals in higher socioeconomic strata generally do not have a dependence on state-provided services and are thus able to stay outside this funnel. They are therefore "invisible" and under-represented in this data.
Units of Analysis: Aggregated Data
1. The Modifiable Areal Unit Problem (MAUP)
From the GIS Dictionary (Association for Geographic Information, 1999):
The segregation of real world features into study areas with artificial boundaries gives rise to the problems of zoning and scale. Depending on the zone and scale used results may differ markedly.
A problem "intrinsic to the analysis of geographical data" (ibid), MAUP continually resurfaced in this analysis. This is problematic because it assumes that values for a large SPA-level unit apply equally throughout the SPA. This is not the case. The West SPA, for example, includes census block groups at both -1.25 to -0.75 standard deviations and >2.75 standard deviations from the median income, i.e., neighbourhoods at opposite ends of the income continuum. Aggregate data is simply not representative of the conditions experienced by the smaller parts making up the whole.
2. Aggregation Effect
Particularly, we observed the aggregation effect, which occurs when small areas are grouped into larger areas - such as when census blocks groups are grouped into large Service Planning Areas (SPAs). In order to compare health data (collected for SPAs) to census data (collected by block groups), values for the variable being considered (income and age, in the preceding cases) were averaged from all block groups within a given SPA. The average was then applied to the entire SPA. This yields a value for the SPA which, though technically correct, is not useful in that it does not accurately represent the census block groups within the SPA.
3. Large Regional Size; Small Sample Size
Trend identification at the SPA level is difficult because it offers such a small sample size. It is difficult to establish a trend with any certainty when there are only eight events from which to make a conclusion.
Drug-Related Arrest Data
The Validity of Arrest Statistics
A major question about drug crime data concerns its ability to accurately represent the distribution of drug use itself. Obviously, drug-related crime is just one indicator of drug use. However, for two major reasons, the potential for drug-related arrests is highest in areas of low socioeconomic status.
The first reason pertains to visibility. Central Los Angeles, with its low median income, has a high rate of homelessness. As a result, drug use is forced onto streets and other public areas with high visibility to police patrols. Conversely, residents in private, wooded estates in Hollywood Hills and Beverly Hills have large amounts of private property to act as a buffer between themselves and the public realm. Drug use can be contained and hidden within the privacy of homes, which may be difficult for police to enter. In upper-class neighbourhoods, recreational and/or chronic drug use can be made nearly invisible to police, and the potential for arrests is much lower.
The second type of statistical bias is a product of the unequal concentration of police patrols. Police departments tend to assign a higher number of officers to the central business district and highly urbanized areas with pre-existing crime problems. Additionally, various departments of the U.S. government have established statistical links between substance abuse and other types of crime. Government policy may compel authorities to "crack down" on drug abuse in neighbourhoods where other types of crime are prevalent. Therefore, it can be argued that drug users in the Central community are more likely to encounter (and be arrested by) a police officer than drug users in largely suburban communities.
The two factors outlined here combine to systematically under-represent drug users in upper-class, suburban communities. This misrepresentation reinforces stereotypes linking drug use and poverty.
Thus, numbers of drug-related arrests presented here cannot be read as valid indicators of any broader variable. They do not accurately quantify drug activity in each police community.
Units of Analysis: Police Communities
1. Inaccuracies in Community Borders
Because GIS-compatible maps of the police communities were not available, we were forced to digitize these regions. The convoluted, jagged edges of the police communities did not lend themselves to manual drawing. Instead, they tended to align with the edges of census block groups, so we defined their borders in this way. In a small number of cases, the census block group borders did not match the police community borders exactly, and we were forced to compromise by placing the entire census block group within one police community.
2. The Modifiable Areal Unit Problem (MAUP) and Agglomeration Effect
As in the analysis of data at the SPA level, MAUP represents a challenge to the accuracy of police community-level data. Statistics at the police community level are assumed to apply equally throughout the community. However, these statistics have been aggregated (and averaged) from a large number of census block groups with widely varying values. Although the compact, community level of analysis is less problematic than the immense SPA regions, there are still cases where a single community is comprised of two or more neighbourhoods of different socioeconomic classes.