Context: illicit trafficking in small arms and light weapons has frequently been cited as a key facilitator of acts of criminal and political violence. However, the nature of arms trafficking remains poorly understood, especially when it occurs in developing countries and outside high-profile conflict zones.
Methods: This paper describes the methods used in the development of the Illicit Small Arms Seizures dataset, which is an in-progress, open source, machine coded event dataset of seizures of illicit small arms and light weapons that have been reported in press articles. The dataset has been under development by the author since September 2016 and it will leverage an extensive library of over 18,000 news articles collected by the Norwegian Initiative on Small Arms Transfers (NISAT) project at the Peace Research Institute Oslo. The articles describe illicit small arms trafficking, and especially document the types and context of weapons discovered and seized by police and other government agencies.
The author has constructed a progressive supervised learning approach to identify, extract, and classify information from NISAT’s corpus in order to produce a rich event dataset. The unit of observation within the dataset is the seizure of or identification of illicit small arms as a tangible and cohesive event. Other indicator variables include the location of and context surrounding the seizure of small arms (e.g. means of transport used by traffickers, intended recipient, or seizure of other illicit goods); though personal data will not be collected. The initial cross-national dataset will cover the years 2008-2016.
Findings: initial descriptive findings from the dataset will be presented in the paper. Methodological caveats will be discussed, including specific issues relating to machine coding and reliance upon press articles as data sources. The paper concludes with initial policy recommendations on how to better prevent illicit small arms trafficking.
