reflection.+how+do+we+know+about+drug+traffic

Examining the Reliability of DIJIN StatisticsDIJIN (Dirección Central de Policía Judicial, or the Central Directorate of CriminalPolice) is a directorate within Colombia’s Policia Nacional. DIJIN is responsiblefor judicial and criminal investigations to support the administrationof justice; as such, it is the primary repository and disseminator of official homiciderecords. Thus, the homicides in the DIJIN database are those known to,and reported by, regional police command posts. While it is an important sourceof information, the DIJIN database may be, for reasons stated above, unlikelyto reflect true patterns of lethal violence in Colombia.ReliabilityIn order to draw conclusions about changes in the true homicide rate, Spagatand CERAC (undated) and González Peña and Restrepo (2006) rely on therelationships between homicide rates across time and space, as reflected in DIJINstatistics. However, to draw plausible conclusions from the data that theyuse requires the strong assumption that the Colombian Police always register aconstant percentage of all homicides. While this is theoretically possible, it isfar more likely that the rate of registration varies (see above and footnotes [6],[7], [8], [9], [10], [11]): in some periods and places, the police register almost allhomicides, while in other periods and places, they register relatively few. Thisvariation directly affects how many homicides are reported. If it became moredifficult to record homicides for any reason following paramilitary demobilization,then the data would be systematically biased downward, toward a spuriousfinding of decreased lethal violence.In the theory of measurement, “reliability” refers to the ability to obtain thesame (or very similar) results from repeated measurements of the same object,whether that object be the weight of an atom, the standardized test performanceof a student, or the number of murders in Colombia during a specific periodof time. In order to claim that lethal violence is declining based on a singledataset, researchers would need to demonstrate that the dataset were reliablein this formal sense. If reporting rates declined while violence remained high,then higher rates of violence in the real world could appear stable or declining ina single dataset. Such a measurement of homicide would be unreliable, because another sample (an additional, independent data set) would be likely to yield different results.In order to assess the reliability of DIJIN data, we would examine cases of homicide reported in a hypothetical complete data source and determine the extent to which cases reported in the DIJIN database overlapped with those in the complete data. Low match rates 17 would indicate underreporting, while changing match rates would imply low reliability (in the formal sense). However,17 See below in Quantitative Analysis. “Match rate” refers to the extent to which cases reported in two data sources overlap. More specifically, it is reported below (see Tables 1 and 2) as the percentage of homicides sampled from an independent data source that are also reported in the DIJIN database. Bibliography:http://www.hrdag.org/resources/publications/CO-PN-CCJ-match-working-paper.pdf