Chapter 1 Introduction

1.1 Introduction

With the ever-increasing advancements in weapons technology, the illicit arms trade has steadily become a greater threat to international security. The small arms trade, consisting of portable weapons and their parts, is not only a profitable good but also a method of gaining power through violent and threatening means. The United Nations Office on Drugs and Crime (UNODC) notes multiple ways in which nations and criminal groups end up with illicit arms, such as fake export arrangements and incorrect recipients. The UNODC notes in Legal Firearms Market that firearms transfers must be documented alongside specifications of what is being shipped and the target country (UNODC, 2019). However, these arms transfers are sometimes labeled instead as machine parts and lead to the underreporting the true number of arms that are being transferred, or they are sent to a different nation than the one reported (UNODC, 2019). The United Nations estimates that the value of the illegal small arms trade may run into the billions, but a majority of the current literature on the small arms trade has failed to account for these unreported arms in their data (UNODA, n.d). The Illicit small arms trade encompasses criminal groups, non-state actors, and nations; however, this paper will be focusing on illicit small arms trades between nations. It is important to note the emphasis on small arms rather than all types of arms trading. The data on large arms trades are considered relatively accurate and complete, given the difficulty of covertly transporting weapons of such magnitude. Illicitly trading a box of handguns presents fewer transportation challenges than illicitly trading a tank, missile, or submarine.

During the conflict in former Yugoslavia in 1992, there was an arms-embargo placed on all the Yugoslav states by the United Nations. Bosnia, Croatia, and Kosovo had to engage with other nations and non-state actors in order to import weapons. Their main suppliers were Iran and Chile who violated the embargo and sent the weapons disguised as other goods or specified the incorrect recipients. A shipment of arms from Chile to Yugoslav states was disguised as humanitarian aid resources for Sri Lanka (Arsovska, 2014).This is an example where nations cheated on their obligation to accurately and honestly report their arms trades. It demonstrates the ways in which nations underreporting their arms trades can impact entire populations and hinder our ability to understand the consequences and causes of the arms trade.

Despite the large impacts illicit arms trades have on nations and individuals, the political science literature addressing the illegal arms trade or attempting to predict or correct for illegal arms trading is scarce. The research within the arms literature has been narrowly focused on particular regions or time periods out of concern for inaccurate data. This has greatly impacted both the quantity and quality of literature around this integral aspect of international relations. Having accurate information on arms trading is essential in order for policy makers to make well informed decisions and assess safety concerns in nations where governments are trading illegally. These concerns and questions have guided our research questions: how do we recognize corrupted network data and how does corrupted network data impact our statistical analyses?

1.2 Corrupted Data and the Arms Trade

Corrupted network data, specifically in the context of illegal arms trading, typically presents itself in two forms. The first is the failure to report any weapons transfers. Data points in this context are simply represented as “NA”. The second is reporting no weapons transfers or a fewer number of transfers than actually took place. The presence of NAs in an arms trading network typically arises where there is a lack of infrastructure that allows such things to be reported. Currently there are multiple methods that deal with missing at random data that have shown to be very effective. Multiple Imputation by Chained Equations (MICE), has risen as one of the most popular methods for addressing missing data. MICE assumes that the values in the data are missing at random, or that the missing value only depends on observed values and is not related to other unobserved values. A conditional distribution is placed upon the variables in the data, and through Gibbs sampling datasets are created from these distributions (Buuren, 1999). In our analysis we are not implementing MICE, but rather setting up a full generative model and sampling from it; these are parallel problems. However, the main issue arises when countries are underreporting their arms trading, resulting in corrupted data.

Nations who are under an arms embargo or who may not want to report all of their arms trading due to possible conflicts of interest, may under report the number of weapons they bought or sold. Thus, while there are clear violations of arms embargos, there can be even greater complications when weapons are exported to states that are geographically close to an embargoed state, to then be transported to that embargoed country (Rothe & Collins, 2011). Thus, states engaging in illicit arms trading will under report or not report at all the number of weapons they are truly trading. Arms databases, in particular the Norwegian Initiative for Small Arms Trade (NISAT), while providing very rich sources of information, are not capturing all arms trades between nations. The NISAT database contains information on arms transfers between states and regions, the types of weapons traded, and the source of the information from 1962 to 2015.

Arms trade takes the form of network data. Network data presents itself in the form of a matrix, sometimes called a socio-matrix, where there are nations (nodes) that are connected to each other through arms trading (edges). While previous work has been done to address data with an abundance of 0s, such as contaminations models including zero-inflated Poisson, there has been a gap in the literature when it comes to data in network format. Thus, this analysis aims to demonstrate how this corrupted network data impacts our statistical analyses and how we can recognize data that may be underreported.