The Tree That Hides the Forest? Testing the Construct Validity of ViCLAS through an Empirical Study of Missing Data
Julien Chopin, Aebi Marcelo · Zenodo (CERN European Organization for Nuclear Research) · 2017
This article tries to identify the reasons that explain the lack of data for some of the variables included in the police databases that are used by crime analysts to establish links between offences in order to detect serial offenders. It is based on an empirical analysis of the missing data in the cases introduced in the French Violent Crime Linkage Analysis System (ViCLAS) from 2006 to 2014. The findings show that the missing data are not randomly distributed, but vary according to the type of variable studied. The highest percentages of missing data are found in variables that refer to behavioural, physical, and distinctive characteristics; while the lowest percentages are found in variables that refer to sociodemographic, descriptive, and situational characteristics. The percentage of missing data increases in parallel with the level of complexity and subjectivity of the data requested, which raises questions about the construct validity of ViCLAS.