This appendix briefly describes the selection and coding of the dependent and independent variables in the Arms Trade Data Set as well as the interview sources and citations. There are two dependent variables: MCW transfers and SALW transfers. The independent variable of interest is importer states’ human rights records. The control variables have been chosen to reflect factors that might have a confounding effect on the relationship of interest (arms transfers and human rights) based on the existing theoretical and empirical literature. The Arms Trade Data Set is significantly broader than previous statistical research in terms of the number of supplier states it covers and its consideration of both small and major conventional arms. It therefore seeks to be as comprehensive as possible despite data problems inherent in this field, which has long had to deal with difficulties of measurement and state secrecy that has only recently begun to fade.
COUNTRIES
The data set contains exporter–importer dyad-years from 1981 to 2010 for 22 top SALW- and MCW-exporting countries and 189 potential importing countries. The exporters include Austria, Australia, Belgium, Bulgaria, Canada, China, the Czech Republic (Czechoslovakia), France, Germany (West Germany), the Netherlands, Norway, Israel, Italy, Russia (Soviet Union), South Africa, South Korea, Spain, Sweden, Switzerland, Turkey, the United Kingdom, and the United States.
The subset of ATT supporter states includes all exporter states that voted in favor of the ATT when it passed the UN General Assembly on April 2, 2013. It excludes Russia and China, which formally abstained from voting.1
DEPENDENT VARIABLES
Because the political attention to, use of, and data for SALW and MCW transfers are dissimilar, the data set contains separate dependent variables for each. Most analyses cover only trade in MCW—where more historical data are available, transparency is more established, and the weapons themselves are easier to track. However, the most work to build multilateral standards and controls in the past decade has taken place with SALW. The ATT process has expanded international discussions on SALW export controls to include MCW. As I describe later in this appendix, differences between data sources also necessitate coding separate variables for methodological reasons.2
Major Conventional Weapons
MCW data are significantly better developed than SALW data. In particular, SIPRI compiles and updates worldwide annual data based on publicly available sources going back to 1950. SIPRI defines MCW as large weapons with a military purpose, covering the following nine categories: aircraft, armored vehicles, artillery, sensors, air defense systems, missiles, ships, engines, or an “other” fulfilling certain qualifications (SIPRI 2007:428–29). Although limited to MCW, SIPRI “provides the most painstakingly researched database” available (Brzoska and Pearson 1994:20) and is an established source in the arms trade literature.3 Since the 1990s, SIPRI data collection has benefited from voluntary state-created transparency initiatives such as the UN Register and the Wassenaar Arrangement.
The dependent variable for MCW transfers (mctransfer) uses SIPRI trendindicator values (TIV),4 which are aggregate dollar figures measuring the core price and value as military resources of actual weapons deliveries within an exporter–importer dyad in a given year (SIPRI 2007:429).5 TIVs are based on an assessment of “the technical parameters of weapons” transferred in a dyad-year. The value represents quantity and quality assigned from “an index that reflects its value as a military resource in relation to other weapons.” (SIPRI 2007:429). Although TIVs consider core weapon prices where that information is available,6 they are not a record of payments made.7 Because transfers can include gifts or aid, in addition to sales, by way of a multitude of financing methods—including barter, discounts, credit, and cash—the standardized TIV measure is substantially more useful in comparing transfers from year to year and country to country (Brzoska 2004; SIPRI 2007). Moreover, it can better factor in transfers of secondhand equipment, to which it is otherwise difficult to assign a monetary value (Durch 2000:8). Finally, TIVs are a uniform measures across countries over time and are updated as new information becomes available, including as past reports are declassified. Thus, although TIV data cannot be combined with other sources, they are an extremely useful tool for research.
SIPRI consults a wide range of public sources in collecting its data: newspapers, books and reference works, official national and international documents, and periodicals and journals. Some records may require an informed estimate by the researcher and tend to err on the side of being conservative (SIPRI 2007:430). It is important to acknowledge that with respect to public sources, an absence of a recorded transfer in a dyad-year can indicate either “no transfer” or a “so far undetected transfer.” However, SIPRI provides the most thorough information available for the global MCW trade.
Small Arms and Light Weapons
The variable for small arms transfers (satransfer) is based on raw records compiled by the Norwegian Initiative on Small Arms Transfers (NISAT; see NISAT 2006).8 NISAT uses the UN definition of SALW, labeling as small arms “those weapons designed for personal use” and as light weapons “those designed for use by several persons serving as a crew” (UN 1997:11). These categories include, for example, revolvers, machine guns, rifles, and ammunition and explosives (UN 1997:11–12). NISAT collects UN Comtrade data9 as well as available national and regional reports and press research but suffers from the lack of a universal system of reporting and methodology (Marsh 2005:2). As a result, data are not streamlined into uniform definitions and measures of quantity or quality. Moreover, multiple reports of a single transfer may exist but cannot be identified as repeat submissions with any certainty, as I explain later in more detail.
Despite the coding difficulties that arise when states do report small arms data, simply tracking the small arms trade is inherently more problematic without their support. As Ian Anthony states, “The rapid and unbroken increase in the movement of standard-size closed containers through ports in countries with a domestic small-arms industry underlines that monitoring of the legal trade can only be done with the consent and cooperation of governments and industry” (1994:34), which have recently become more common (Holtom 2008). In its first yearbook, SAS called small arms transfers “a unique kind of terra incognita” that “remain statistically primitive and underdeveloped” (2001:61). Since then, however, numerous national, regional, and international sources have appeared, helping generate a broad sketch of the small arms trade.
Because of the rough nature of SALW data, satransfer is dichotomous: the number 1 indicates the presence of a transfer in a dyad-year, and zero indicates no record of a transfer in a dyad-year. Three significant data problems necessitate a dichotomous SALW variable. First, the absence of a transfer record does not mean with certainty that no transfer took place—only that one was not listed in the numerous sources used to compile the database. Without completely open government records for all years, it is simply impossible to distinguish a missing variable from a “no transfer” record. This is especially problematic with regard to data from the Cold War, which Nicholas Marsh refers to as the “heyday of gray market arms transfers” (2002:221). The superpowers delivered arms to conflicts worldwide in an attempt “to subvert the global influence of the other by clandestinely arming the enemies of their enemies” (221). Rather, it can only be assumed that the figures are generally underestimates, especially during the Cold War.
Second, one record in a dyad-year may list price and not volume, but another may list volume and not price. The specific number of weapons is also often missing. This disparity makes it difficult to create a variable based on either price or quantity because only partial information is usually available. As SAS points out, “Many governments report only the value or tonnage given in their customs receipts, leaving the actual number of weapons to guesswork” (2003:99). At the same time, the values of transfers—if given—are not necessarily accurate or useful because states may acquire arms through bartering, credit, or gifts as well as cash, which a raw price report does not capture (Brzoska 2004:113; Levine et al. 1997).
Finally, a single dyad-year may list more than one record of a single transfer. As a rule, NISAT utilizes multiple sources (“mirror statistics”) in compiling its database and lists every record found for a dyad-year. For example, there may be at least two sources of data on a transfer, one taken from exporting-country reports and the other from importing-country reports (Marsh 2005). This approach helps to broaden the picture of the SALW trade, particularly for exporters or importers who do not regularly submit reports. However, it is impossible to determine whether the records reflect multiple sources on a single transfer or are indeed multiple transfers. Exporters and importers may provide either different types of information (e.g., price or quantity) or even slightly (or grossly) different values for the same type of information about the same transfer. Data aggregation is therefore unwise, and NISAT explicitly advises against it because of the nonstandardized nature of the raw data.
As a consequence, although it is possible to identify whether a small arms transfer has taken place, it is impossible to provide accurate information about the price paid or the amount transferred. Without improved records, satransfer is therefore most responsibly coded dichotomously. At the same time, it remains a crucial component of this analysis. Given the dearth of political and scholarly attention to small arms until recently, there is limited knowledge of states’ patterns of exports in this area. This gap widens when it comes to quantitative analyses and suggests that using available data with an eye informed as to its shortcomings is well worth the effort. The value-added of including satransfer is high, not only for the research at hand—small arms have been at the center of policy discussions and the conduit for promoting “responsible” export controls—but also for arms control research more broadly, which lacks a fundamental understanding of states’ small arms export practices.
INDEPENDENT VARIABLES
The statistical model explores the relationship between arms transfers and recipients’ human rights records over time and contains four additional control variables: GDP per capita as a measure of development, democracy, internal conflict, and oil production. Table B.1 summarizes the independent variables. Each of the control variables is included because it may not only have an effect on arms transfers but also have an effect on human rights conditions in recipient states. As a result, variables such as alliance—which do tend to have a large significant effect on arms transfers but are not theoretically or empirically linked to recipient human rights—are excluded from the model.
According to empirical research, the presence of both internal conflict10 and oil production11 worsens a state’s human rights. Democracy12 and more advanced economic development,13 in contrast, are associated with better human rights. Omitting one of these variables therefore has the potential to skew the results. For example, because GDP per capita is affiliated with both positive human rights and arms transfers, its exclusion from the model risks results that overestimate the relationship between good human rights and arms transfers.
Human Rights
Arms importer human rights records are the independent variable of interest. Human rights are at the heart of evolving humanitarian standards of arms export controls, set out in codes of conduct and other statements of “responsible” arms transfer controls. Yet whether recipients’ human rights matter in arms transfer decision making and, if so, how they matter are open questions. Researchers have debated the definition, measurement, and quality of human rights data.14 Most data sources focus on physical integrity rights based on annual reports published by the U.S. DOS and Amnesty International. I use the Political Terror Scale (PTS), which ranks states annually based on these reports and provides a score from 1 to 5 for each country and each year in the data set beginning with 1980 (Gibney and Dalton 1996).
TABLE B.1. INDEPENDENT VARIABLES
| Independent variable |
Source |
Coding |
Justification |
| Human rights |
PTS |
1 (rare or extremely exceptional human rights violations) to 5 (frequent and severe violations extended to whole population) |
Included in new arms export criteria but violations may increase demand for weapons |
| GDP per capita |
UN |
Estimated from the national account aggregates and logged |
May affect recipients’ human rights records; greater wealth expands resources to purchase arms |
| Democracy |
Polity IV |
−10 (autocracy) to +10 (democracy) |
May affect arms exports from democratic exporters; may also affect recipients’ human rights. |
| Oil production |
Centripetalism |
Millions of barrels of oil produced per day per capita |
Provides producers with resources with which to purchase arms and favored arms export treatment from arms suppliers |
| Internal conflict |
Uppsala/PRIO |
0 (no conflict) to 2 (full war) based on number of battle-related deaths |
Increases demand for weapons, affects recipients’ human rights |
PTS coding rules are detailed in Gibney and Dalton (1996), starting with level 1 for countries “under a secure rule of law” where political murder is “extraordinarily rare,” torture is “rare or exceptional,” and “people are not imprisoned for their views” (73). At the other end of the scale, level 5 is for countries in which leaders “place no limits on the means or thoroughness with which they pursue personal or ideological goals” and “murders, disappearances, and torture are a common part of life” for the whole population (74). I create dummy variables for each level of the PTS scale, ranging from “very good” to “very bad.” Because the PTS is ordinal rather than continuous, this treatment of the data is more accurate and technically correct:15 there is no reason to assume that the difference between level 1 and level 2 is the same as the difference between level 2 and level 3, and so on (Wooldridge 2000:221–24). The level of human rights score with no violations (1 or “very good”) is removed from the statistical analysis as the reference category for the four remaining dummy variables. Because this category is dominated by wealthy democracies—which, as typically “very good” human rights performers, are less interesting to the study here—this choice does not detract from the analysis.
Like any human rights data, PTS data are “inherently subjective,” yet, as the Human Security Centre notes, “[it] sheds much-needed light on a murky corner of human insecurity” (2005:79). PTS also is the most comprehensive scale in terms of years and countries and includes separate variables for U.S. DOS and Amnesty reports (Poe, Carey, and Vazquez 2001).16 Given the focus of the analysis on government arms export decision making, I use primarily the DOS-coded data, which cover more countries than the Amnesty data and provide some insight into government perceptions of human rights performance.17 Of course, with both of these sources, political bias may be present. Joe Foweraker and Roman Krznaric (2000) find that DOS reports may be biased against left-wing governments and Amnesty reports in favor of them. Steven Poe, Sabine Carey, and Tanya Vazquez similarly observe that in the early years of such reporting in particular, “the US … tended to be somewhat less harsh than Amnesty in evaluating the human rights practices of other governments” (2001:661).18 Yet they also point out that only a low proportion of variance between the sources is explained by these biases, concluding that “we have absolutely no reason to believe that the vast majority of the differences between the reports are systematic” (670).
Gross Domestic Product per Capita
This variable, which is logged in the analysis, comes from the National Accounts Main Aggregates Database maintained by the United Nations Statistics Division (2006). GDP per capita provides a broader measure of a recipient country’s wealth and is an indicator of potential resources available for arms procurement. A profit-oriented producer would presumably have a greater interest in trading with a wealthier country. More advanced economic development is also associated with better human rights.
Democracy
Democracy is an important component of the analysis: democratic importers may be more likely to adhere to international rules and norms, including greater respect for human rights.19 Moreover, it is also thought that democracies share a closer relationship with each other in the international community, suggesting that arms trade between democracies would be higher. The variable polity2, taken from the Polity IV data set (Marshall and Jaggers 2005a, 2005b), indicates the level of democracy in the recipient state. Its scale of government types ranges from strongly autocratic (−10) to strongly democratic (+10) based on weighted assessments of criteria including competitiveness of political participation and executive recruitment, openness of executive recruitment, and constraints on the chief executive. Although there is a long-standing debate about how democracy is best defined and measured,20 Polity provides the most widely accepted and most comprehensive data available on democracy (Munck and Verkuilen 2002).
Oil Production
Oil production is included in the analysis as a single variable taken from John Gerring, Strom Thacker, and Carola Moreno’s (2005) Centripetalism data set, which provides data on millions of barrels of oil produced per day per capita.21 Existing research suggests that major oil producers are privileged recipients on the arms market. This was particularly the case during the oil crisis of the 1970s, which increased the resources available to oil-producing states to buy arms and the desire of arms supplier states to sell them.22 This relationship has endured beyond the Cold War as an influential factor in the foreign policies of major powers in the Middle East for three key reasons (Chapman and Khanna 2006; Prados 2002:9). First, oil is a lucrative resource. It can raise the income of producing states and thus increase their ability to purchase arms (Brzoska and Ohlson 1987; Pearson 1994). Second, due to insecurity and border-control issues, oil-producing states in the Persian Gulf in particular may exhibit a high demand for arms (Chapman and Khanna 2006). Finally, oil is thought to be a significant reason that major powers have sought favorable relations in the Middle East—often, it is alleged, at the expense of human rights, democratic values, and regional stability.23 Arms transfers have long been an important means of cultivating good relationships with interested countries.
Internal Conflict
Since the end of the Cold War, the vast majority of conflicts have taken place within states and not between them. An internal conflict is defined as a conflict between a state’s government and internal opposition groups, without external intervention (Gleditsch et al. 2002). Conflict increases the demand for arms. Small arms in particular are seen as a primary tool of internal conflict, although states have also supplied MCW to government and rebel allies to help tip the scale in a conflict. States engaged in internal conflicts are also likely to experience higher levels of human rights violations.
The widely used Uppsala/PRIO (Peace Research Institute Oslo) Armed Conflict Dataset (Gleditsch et al. 2002) includes both high- and low-intensity conflict at annual death thresholds lower than Correlates of War data (Human Security Centre 2005:18–20).24 The value assigned to conflict intensity is based on the number of battle-related deaths each year: 0 (no conflict); 1 (minor; at least twenty-five military or civilian battle-related deaths in a year); and 2 (war; a minimum of one thousand battle-related deaths per year).25 For the same methodological reasons just described with the human rights data, I use dummy variables for high- and low-level internal conflict in the analysis, with “no conflict” as the reference category.26
INTERVIEWS
The interview citations are designed to maintain the interview subjects’ anonymity. Due to the sensitive political nature of the topics under discussion, all but a few subjects agreed that their general affiliation but not their name or specific agency could be used as a public reference.27 Because the arms trade community is small and relatively tight-knit, I have refrained from listing any specific government agencies, defense companies or associations, or civil society groups, even where respondents permitted it, because doing so could make others more easily identifiable. Each interview subject is therefore coded by a number comprising the interview number; the subject’s general affiliation (1 for government official; 2 for NGO; 3 for industry; 4 for other, such as expert); the interview year; and the Correlates of War country code for the subject’s national expertise/affiliation. For example, an interview number might be 60108220 (60 = interview number, 1 = affiliation, 08 = 2008, 220 = country code).
TABLE B.2. CASE STUDY INTERVIEWS
| Country |
Correlates of war code |
Number of participants |
| Belgium |
211 |
18 |
| Brazil |
140 |
2 |
| France |
220 |
6 |
| Germany |
255 |
15 |
| Switzerland |
225 |
9 |
| United Kingdom |
200 |
12 |
| United States |
002 |
5 |
Note: Participants in Switzerland were members of the international NGO and research community based in Geneva.