WPS4202 POST-CONFLICT TRANSITIONS WORKING PAPER NO. 10 WEAPONOMICS: THE GLOBAL MARKET FOR ASSAULT RIFLES Phillip Killicoat Department of Economics Oxford University This paper introduces the first effort to quantitatively document the small arms market by collating field reports and journalist accounts to produce a cross-country time-series price index of Kalashnikov assault rifles. A model of the small arms market is developed and empirically estimated to identify the key determinants of assault rifle prices. Variables which proxy the effective height of trade barriers for illicit trade are consistently significant in determining weapon price variation. When controlling for other factors, the collapse of the Soviet Union does not have as large an impact on weapon prices as is generally believed. Key words: small arms, Kalashnikov prices, black market World Bank Policy Research Working Paper 4202, April 2007 The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Policy Research Working Papers are available online at http://econ.worldbank.org. 1 INTRODUCTION Small arms are estimated to be responsible for between 200,000 - 400,000 deaths around the world each year. Approximately 20,000 ­ 100,000 of these firearm deaths occur in conflict settings (Small Arms Survey 2005, Kopel, Gallant and Eisen 2004, and Lacina and Gleditsch 2005). As economic commodities, firearms are subject to the forces of demand and supply and are actively traded on legal and illicit markets. The small arms market may be viewed as a function of the incentives and constraints faced by buyers, suppliers and regulators. This paper introduces cross-country, time-series data on assault rifle prices thus making it possible to quantitatively examine the nature of the small arms market. Small arms are attractive tools of violence for several reasons. They are widely available, low in cost, extremely lethal, simple to use, durable, highly portable, easily concealed, and possess legitimate military, police, and civilian uses. As a result they are present in virtually every society. (Boutwell and Klare 1999) Despite being a key component in conflict, small arms have only recently begun to receive academic attention. So far research has been almost exclusively case-study driven making it difficult to draw general empirical lessons. Book length treatments of small arms which follow this trend include Boutwell and Klare (1999) and Lumpe (2002). Brauer (2007) surveys the small arms literature in the forthcoming Handbook of Defense Economics and concludes that the small arms market has not been well examined theoretically, or empirically. The first tentative steps towards generalizable models of the small arms market are currently underway. Brauer and Muggah (2006) develop a conceptual theory of small arms demand as a function of means and motivation, an adaptation of the standard determinants (income, prices and preferences) of neoclassical consumer demand theory (Varian 1992). On the supply side, Marsh (2007) develops a conceptual model for the illicit acquisition of small arms by rebel groups. Among other hypotheses, Marsh's model predicts that the more liquid is 2 the arms supply in a particular country, i.e. the more easily individual combatants can obtain weapons through independent suppliers, the more difficult it will be to mount and maintain a united and coordinated insurgency. There are a number of reasons why small arms have been all but ignored in the quantitative analysis of conflict. The historic state-centric bias of defense economics led to an almost exclusive focus on inter-state military strategy. In relation to military weapons, research has principally been concerned with the development and acquisition of large-scale military technology, such as nuclear weapons. Perhaps the most important reason for the dearth of attention given to the role of weapons in civil war is that usable data have been unavailable. The policy research community, led by the Small Arms Survey (SAS), the UN's Small Arms and Demobilisation Unit, the Bonn International Center for Conversion, and the Norwegian Initiative on Small Arms Transfers (NISAT), has produced a great deal of survey and case-study work. However, no statistical analysis of the growing volume of survey information has yet taken place. DATA Existing data on aspects of the small arms market are extremely limited. Since 2001, the Small Arms Survey has gathered a range of information on small arms products, stockpiles, producers and trade. Despite occasional references to observed prices, the Survey has not regularly collected price data which would be of most benefit for generating inferential statistics. Collecting price data for panel analysis requires an operational definition of the variable of interest that will provide consistency across time and countries. In the case of small arms there is an obvious choice: the AK-47 assault rifle. Of the estimated 500 million firearms worldwide, approximately 100 million belong to the Kalashnikov family, three-quarters of which are AK- 47s (Small Arms Survey 2004). The pervasiveness of this weapon may be explained in large part by its simplicity. The AK-47 was initially designed for ease of operation and repair by glove-wearing Soviet soldiers in arctic 3 conditions. Its breathtaking simplicity means that it can also be operated by child soldiers in the African desert. Kalashnikovs are a weapon of choice for armed forces and non-state actors alike. They are to be found in the arsenals of armed and special forces of more than 80 countries. In practically every theatre of insurgency or guerrilla combat a Kalashnikov will be found. The popularity of the AK-47 is accentuated by the view that it was a necessary tool to remove colonial rulers in Africa and Asia. Indeed, an image of the rifle appears on the Mozambique national flag, and "Kalash", an abbreviation of Kalashnikov, is a common boy's name in some African countries. The AK-47's popularity is generally attributed to its functional characteristics; ease of operation, robustness to mistreatment and negligible failure rate. The weapon's weaknesses - it is considerably less accurate, less safe for users, and has a smaller range than equivalently calibrated weapons - are usually overlooked, or considered to be less important than the benefits of its simplicity. But other assault rifles are approximately as simple to manage, yet they have not experienced the soaring popularity of the Kalashnikov. The AK-47's ubiquity could alternatively be explained as a result of a path dependent process. Economic historians recognize that an inferior product may persist when a small but early advantage becomes large over time and builds up a legacy that makes switching costly (David 1975). In the case of the AK-47 that early advantage may be that as a Soviet invention it was not subject to patent and so could be freely copied. Furthermore, large caches of these weapons were freely distributed to regimes and rebels sympathetic to the Soviet Union - more freely, that is, than weapons were distributed by the US - thereby giving the AK-47 a foothold advantage in the emerging post-World War II market for small arms. According to a path dependence interpretation, inferior durable capital equipment may remain in use because the fixed costs are already sunk, while variable costs (e.g. ammunition, learning costs for new recruits) are lower than the total costs of replacing Kalashnikovs with a new generation of weapons of apparently superior quality. Whatever the exact causes, it remains that for the last half-century the AK-47 has enjoyed a near dominant role in the market for assault rifles making it the most persistent piece of modern military technology. Since the technology 4 used in the AK-47 is essentially unchanged from the original, one may be confident that the prices observed across time and countries are determined market conditions rather than changes in the product. Data Sources The weapon price data are compiled from a range of journalistic reports and industry interviews. The unit of analysis is the price in $US for each country for each five-year period for a non- government entity to take possession of an AK-47 assault rifle. The foundation of the dataset was generated with the assistance of the Small Arms Black Market Archive, maintained by the Norwegian Institute for Small Arms Transfers (NISAT 2006). The Archive contains over 9,000 documents relating to illicit small arms trade. Articles with references to quoted prices or reported transactions involving AK-47 or equivalent assault rifles were extracted and the information converted into the data format using the coding rules outlined in Appendix A. References to assault rifle prices were extracted from the back editions of the Small Arms Survey, which have been obtained on an ad hoc basis from field work. The dataset also benefited from interviews with arms industry experts who have had considerable experience with arms bazaars throughout Africa and Asia. Of particular note is Brian Thomas, an investigative journalist, who has been following the illicit arms trade from factory-to-fight for the last 15 years and has assiduously recorded the going prices for assault rifles in a range of locations at different times. The frequency distribution of data sources for price observations is as follows: NISAT Small Arms Black Market Archive (58%); Small Arms Survey (17%); US Alcohol Tobacco and Firearms Authority (16%); Brian Thomas (6%); other sources (3%). Summary of Kalashnikov Price Data This section discusses the strengths and weaknesses of the data, and presents descriptive summary statistics. The major strengths of the data include the broad coverage of countries for which at least one data point was obtained (117); a consistent operational definition of the price variable across time and countries; collection of multiple country-period observations to verify 5 that data is of the correct order of magnitude. Furthermore, the AK-47 price variable may be considered a strong proxy for the price of conflict-specific capital. A potential weakness of the data relates to the randomness of the sample collected. The time dimension suffers from a temporal selection bias. There are relatively more observations for more recent periods. For the period 1986 to 1990 there are 46 unique country observations, whereas for 2001 to 2005 there are 101. This is most likely a due to the combination of more thorough information dissemination facilitated by the internet and the recent increase in attention given to the small arms trade. The country dimension potentially involves a nonrandom sample as there are relatively more weapon price observations for low-income countries which have experienced civil war compared with peaceful low-income countries. Small arms will tend to be more actively traded in or near war-affected countries. A concern is that journalistic accounts may exaggerate or only report extreme prices. One would expect such measurement error to be biased downwards in poor or war-affected countries. Adherence to the coding rules above generally precludes extreme or outlier data points as they do not conform to the definition which is used to provide a consistent measure of equivalent AK-47 trades. Summary Statistics The dataset potentially contains i = 208 countries over t = 4 time periods. The 208 countries are those for which the World Bank collects data for the World Development Indicators (WDI) data base. Subtracting data points for those countries which did not exist due to achieving independence later than 1986 leaves 742 potential observations. As shown in Table I there are 335 independent country-period data points for weapon prices. Coverage for just under half of all potential data points would suggest sufficient coverage for purposes of inferential statistics. In addition to a temporal selection bias towards the present, there are comparatively more observations for Africa and the Middle East, and fewer in Western Europe. The low rate of observation in Western Europe (12 observations in the whole sample) may give rise to sample 6 selection effects which must be addressed in the future. One possible method to overcome this would be to impute AK-47 prices from the prices of competing, equivalent assault rifles. Figures 1 and 2 track the movement of average weapon prices for regions, and for countries with civil conflict experience. What can be seen is that in peaceful and developed countries weapon prices have been rising. In conflict-affected countries prices has remained roughly constant while in Africa prices have in fact been trending down. A country is deemed conflict-affected if it has experienced a civil war in the last 20 years. THE SMALL ARMS MARKET This section develops a model of the small arms market based on a simultaneous equations model of demand and supply. Demand for small arms depends on their relative price (P), income (I) and the motivation for owning a weapon (M). The supply side of the small arms market is determined by price (P), the prevailing regulations in relation to small arms (R), and intrinsic supply costs (S). The structural demand and supply equations of this simultaneous equation system are given by: Qd = -a - bP + cI + dM (1) Qs = e + fP - gR - hS (2) Setting (1) equal to (2) for an equilibrium: Qs = Qd (3) e + fP - gR - hS = -a - bP + cI + dM (4) Solving these equilibrium conditions for the endogenous dependent variables price (P) and quantity (Q) yields the following reduced form equations: 7 P = - b e + a c + f +b + f d I + b+ f M + b + f g R + b h S (5) + f Q = - be - af cf df b + f +b + f I + b+ f M - b gb S (6) + f R - b hb + f Since we do not currently have country estimates for the quantity of Kalashnikov trades (Qi), it is not possible to estimate both reduced form equations. Hence the structural parameters (a... g) from equations 1 and 2 cannot be empirically estimated. With the benefit of the collected weapon price data we can nevertheless estimate the reduced form equation for weapon price. While the magnitude of the estimated coefficients of the reduced form equations should not be interpreted in the normal linear fashion, their signs and significance can provide meaningful insight into the nature of the small arms market. In order to estimate the reduced form price equation, it is necessary to obtain data for variables which proxy the desired concepts (Income (I), Motivation (M), Regulation (R), Supply costs (S)). Table III outlines the empirically observed variables which will be used to estimate the reduced form price equation. A four-period (20 year) cross-country panel is used to estimate the reduced form model for weapon price determinants: Pit = 0 + 1Iit + 2Mit + 3Rit + 4Sit + eit (7) The estimation method used is random effects generalized least squares (GLS). The random effects approach is appropriate where there is reason to believe that some omitted variables may be constant over time but vary between cases (e.g. geography) which could be managed with a fixed effects estimator, while other omitted variables others may be fixed between cases but vary over time (e.g. illicit supply sources) and would be best served by a between estimator. It is possible to include both types using the random effects estimator which is a weighted average of fixed and between effects estimators (Wooldridge 2002). In order to determine whether random effects provides a consistent estimator, we run a Hausman test against the less efficient but 8 assuredly consistent fixed effects model. The Hausman test for the basic model (column 1 in Table IVa) yields an insignificant -value (0.26) for the null hypothesis that random effects is consistent and efficient relative to fixed effects. Results Table IVa and Table IVb present regressions based on the reduced form weapon price determinants model (Equation 7) for the global sample of weapon prices. Column 1 begins with a single variable for each concept (income, motivation, regulation and supply costs). Subsequent versions test the robustness of the model to alternative specifications of the explanatory variables. Income It is expected that the higher is per capita income (I) the higher will be weapon prices, due to the partial non-tradability of weapons from official trade barriers. Results from alternative variations of the model only weakly support this hypothesis. According to competitive international trade models, free trade will equalize commodity prices. However, non-government weapons trade between countries is almost always contraband. To the extent that laws prohibiting weapons trade are enforced, weapons will take on the attributes of non-tradable goods. The price of this class of good is determined by domestic factor prices, most importantly labor, and labor costs will in general be larger the higher is income. Due to the partial non-tradability of weapons, the theoretically appropriate measure of income is GDP per capita in purchasing power parity (PPP) terms. Other measures of income also find a positive relationship between income and weapon price. However, variables which measure income in nominal or absolute terms are more strongly subject to income's correlation with governance variables. One might expect causation to flow from income to governance: the higher is income the more tax governments have at their disposal to spend on effective regulation and law enforcement. But available evidence suggests that the causal impact of income on governance is negligible, and causation is more robustly demonstrated to operate in the opposite 9 direction (Kaufmann, Kraay and Mastruzzi 2005). When the PPP measure of income was replaced with income in constant US$, the regulatory variable R (government effectiveness) was rendered insignificant. The PPP income measure is less susceptible to correlation with governance indicators and can be more confidently interpreted as the wealth mark-up on weapon prices for a given regulatory environment. Motivation Obtaining a satisfactory proxy for the motivation (M) to purchase assault rifles is a difficult task. In the first instance, income growth is adopted as a measure for the desire to buy weapons. Negative income growth has been found to increase the proneness of a country to civil war outbreak (Collier and Hoeffler 2004), even when accounting for the endogeneity of economic growth in the conflict process (Miguel, Satyanath and Sergenti 2004). It is also found to increase the incidence of violent crime (Fajnzylber, Lederman and Loayza 2002). Therefore, we would expect negative income shocks to lead to an increased motivation to purchase weapons for the purposes of crime or conflict. In the estimated model, the coefficient on lagged income growth is not statistically different from zero (columns 1 and 2). The inconclusiveness of this parameter estimate may be the result of competing effects in the small arms market during economic downturns. While one expects the demand for weapons (for crime and conflict) to drive weapon prices up, it is conceivable that there is an even stronger supply effect. Agents on the margin of the legal labor market become unemployed in an economic downturn and a fraction of those unemployed take on employment in the black market (including the arms trade), which is profitable relative to no work at all. The extra (illicit) employment in arms trade creates a more competitive arms market and the increase in supply may offset the increase in demand. Since the results for lagged income growth are insignificant it is not possible to determine whether the supply or demand effect dominates. A rationalization for the observed parameter estimate of zero is that the illicit weapons market adapts well to changes in economic conditions so that the effect of economic shocks on weapon price is neutralized. 10 Another hypothesized driver of the motivation to purchase assault rifles is civil conflict, the setting where such weapons are mostly likely to be used for their intended purpose. An indicator variable for civil war onset is included to proxy demand for weapons for rebellion. The war start variable is coded one if in a five-year period a civil conflict claims at least 25 deaths in a given year. While the parameter estimate was positive it was insignificant (column 10 in Table IVb) so it is not possible to conclude that on average there is a significant demand side effect on weapon prices during the period of conflict onset. The result was similar for the 1,000 battle death threshold. A range of other variables were additionally tested in an effort to capture the motivation to purchase weapons. The proportion of young men (the demographic group most likely to purchase weapons); the proportion of young men interacted with income growth, and schooling (it is hypothesized that uneducated young men and those who experience negative income shocks are prime candidates for seeking weapons); finally, the average rate of homicide as an approximate measure for the underlying proclivity towards violence in a country was tested. All of these measures for motivation proved insignificant in explaining weapon price. This is not to conclude that motivation is unimportant in determining weapon price. Rather, it may indicate that better measures of preferences for purchasing weapons are required, and that decomposing motivation effects is not something that can be achieved in the basic framework currently under analysis, especially as the parameter estimates are for the reduced form, not the structural demand and supply equations. An alternative explanation for the insignificance of demand side variables is that the price elasticity of supply is very large relative to the price elasticity of demand for assault rifles. This is discussed further in the section on supply costs. Regulatory Effectiveness Almost all countries have legislation designed to control the trade and possession of small arms. What differs is the ability of governments to enforce these laws. We expect that the more effective a government is at upholding its law, the greater will be the cost to trade weapons, legal or otherwise. The regulatory variable (R) is intended to capture the height of the trade barriers that must be overcome in order to sell a weapon. 11 A number of measures of regulatory effectiveness are used and all indicate that better enforceability of laws and regulations raises the price of weapons. The World Bank's government effectiveness variable which measures the competence of the bureaucracy is everywhere positive and significant. Data from the International Country Risk Guide (ICRG 2005) confirms the importance of regulatory capacity as a determinant of weapon price. Democratic accountability measures are significant suggesting that checks on different levels of government and public services are also important in enforcing law in relation to illicit weapons (column 7). The ICRG law and order variable is intended to proxy the on-the-ground ability of police to enforce the law and prosecute weapons violations. The parameter estimate is positive, but less convincing than expected (column 8). This may be explained by a demand-effect at very low levels of law and order. Households and groups are acutely aware when internal security forces are ineffective and may attempt to fill a security vacuum with their own weapons acquisition, whether for self-defense, crime or conflict. The lesser significance of the ICRG variables may be due to their reduced coverage relative to the World Bank's variables. As a check for whether the effect of varying sample sizes are significant, regressions were run with the World Bank governance data on the sample for which there was ICRG data. The results were not significantly different in the smaller samples. The variables used to proxy regulatory effectiveness (R) are all ordinal indicators. Since these variables are not cardinal, the effect of a change from, for example, -1 to 0 is not necessarily commensurate with an improvement from 0 to +1. As such, the parameter estimates cannot be interpreted in the standard linear fashion. In order to verify that the ordinal dimension of these variables is not biasing estimation, segments of the governance variables are pooled together. Dummy variables for each third of the government effectiveness distribution are generated and included in the weapon price regression. In the first instance, the bottom third of countries is included, and the Africa dummy is excluded . The bottom third governance indicator variable is independently significant (column 12), but when Africa is again included (column 14) the Africa dummy maintains its significance and yields a similar parameter estimate, while the segmented 12 governance dummy becomes somewhat less significant ( = 0.12). This procedure was also undertaken for the 20th and 25th percentile segments of the distribution with similar results. Since the remaining parameters are not affected by respecification, it may be concluded that the ordinal properties of the governance variables do not systematically bias the estimates. The regulatory effectiveness variable is concerned with the effective height of the trade barriers that need to be overcome in order to trade a Kalashnikov. The empirical governance variables considered so far account for the relative freedom of within-country trade. Arguably, however, between-country trade barriers are at least as important as within-country barriers. The ideal variable would be some measure of the porousness of a country's border since the vast majority of cross-border small arms transactions are likely to be illicit. Since no such data currently exist it is proposed to use a dummy variable for African countries. Africa provides a natural experiment because its countries on average possess a higher number of neighbors than the rest of the world (3.4 versus 2.1), that are considered to have more porous borders than the rest of the world (CIA 2005). Even controlling for income, government effectiveness, war legacy and supply cost variables, being located in an African country makes purchasing an assault rifle on average over US$200 cheaper than elsewhere. It is postulated that this staggering Africa-discount is predominantly driven by porous borders. Since borders are more porous than elsewhere, the trade in assault rifles across the African continent approaches a deregulated market in which prices converge and there are only negligible trade barriers that arms supply must overcome to meet demand. At any one time, only a few African countries have very high demand for weapons due to conflict. This demand profile across the continent changes over time as localized tensions rise and recede. Porous borders enable the entire supply of weapons on the African continent to meet whichever country currently has high weapons demand. 13 Supply Costs The supply costs variable (S) in the small arms market model is designed to capture the intrinsic non-regulatory costs involved with supplying arms. A range of empirical variables are used to represent the key factors that affect the underlying cost of supplying assault rifles. The supply cost variable that proves most robust is neighbors' average military expenditure. This variable measures the average of neighboring countries' annual government military expenditure as a share of GDP. It is theorized that the strong negative correlation between neighbors' military expenditure and weapon price is driven by spillovers and leakages. Spillovers arise where some fraction of a country's military spending is allocated to supplying arms directly to anti- government forces in rival neighboring countries. The exact reasons for governments supplying foreign rebel forces with arms are not considered here, but one may conjecture that such supply involves some strategic decision designed to destabilize or divert the attention of a threatening neighbor's regime. The leakage effect arises not from a conscious effort by neighbors, but from misappropriation of official weapons stocks by arms dealers and rebels. Such acquisition is typically facilitated by unauthorized sales by defense force personnel (i.e. corruption) or the forcible seizure of weapons stocks during combat or raids on arsenals, which are then sold across borders. Surprisingly, own-country military expenditure was not a satisfactory explanator of weapon price. Indeed, it had the opposite sign to neighbors' military expenditure (column 9). An explanation for this result is that most illicit purchases of weapons will not be from officials to non-government agents of the same nationality. In general, defense forces would not wish to destabilize their own regime by facilitating arms trade with domestic rebels. Even at lower levels within the military, the private incentives of soldiers making some extra money from unauthorized sales to domestic rebels is likely to be outweighed by the expected cost of being caught and dealt corporal or capital punishment. Moreover, there is a deterrent effect of own military expenditure on the feasibility of weapons trade. Where a country has a strong military presence (as proxied by a high level of military expenditure), it would be imprudent for non- government entities to openly trade or parade about with large quantities of conflict-grade weapons. 14 The supply cost variable that seeks to proxy the stock of weapons in circulation is a variable called civil war legacy. The legacy variable is generated using the cumulative civil war battle- deaths since 1960. Since the majority of battle deaths are caused by weapons, the number of battle deaths may be considered a suitable proxy for the quantity of active weapons in a country. In the same way as the magnitude of a war 30 years ago matters proportionately less than an equivalent-sized battle last year, the weapons used to prosecute the war depreciate over time. A discount rate of 5% is applied to recognize depreciation, consistent with a Kalashnikov's life expectancy of up to 50 years. As an approximation of the number of active weapons, the legacy variable is reasonably robust to various model specifications. Its parameter estimate is negative significant conforming with elementary price theory which predicts that, all else equal, the more plentiful is a commodity, the cheaper it will be. To the extent that the legacy variable provides a proxy for the stocks of non-government weapons, it also illustrates why weapon supply is considerably more elastic than demand. According to the basic theory of price elasticity of supply where there are higher stocks, supply agents (weapons traders, rebel groups) will be able to respond to changes in demand relatively more quickly and hence supply will be relatively more elastic. It is commonly believed that the collapse of the Soviet Union released inestimable stocks of weapons onto the world market. This view has been popularized in a recent Hollywood film, Lord of War, where Nicholas Cage plays a Ukrainian arms dealer who profitably liquidates the former Soviet state's military arsenal. According to conventional wisdom, weapons trade during the Cold War was based on political affiliation, but since the collapse of communism it has been driven by profit-seekers. Another way of conceiving this hypothesized transition is in terms of industrial organization: until 1991 there was a duopoly in the weapons market (USA and USSR). Since then the global market has been effectively deregulated with numerous agents operating in a competitive market. Was the collapse of the Soviet Union a significant supply shock for the illicit weapons market? Regression results suggest not. At the very least, it is not as important as previously believed. When controlling for other factors, the coefficient on the dummy for the post-Soviet collapse 15 period is not significant at conventional levels (column 6). This result suggests that the historical case for a structural break in the global market for small arms has been overstated. An explanation for this finding is to be found in the role of secondary markets. Since weapons are durable goods they can, like shares in a firm, be repeatedly sold from agent to agent. During the Cold War, even though the superpowers thought they were giving or selling weapons to their political allies, these weapons were regularly - and profitably - sold on to secondary (or black) markets which had no regard for the political stripe of the initial source of the weapon. Two caveats to this finding should be acknowledged, however. First, there is only one observation period (1986-1990) before the Soviet collapse. Second, there are only 46 observations for the pre-collapse period, whereas there are more than 80 for each of the three subsequent periods (see Table II). While the collapse of the Soviet Union did not in itself appear to be a significant supply shock for the small arms market, the role of the Soviet Union and its successor states as sources of weapons does yield significant parameter estimates. Distance from Moscow is adopted as a proxy for the transport costs of getting weapons (in this case Kalashnikovs) from their initial source to the secondary markets on which they are traded. The distance from Moscow variable is positively correlated with weapon prices for all model specifications indicating that transport costs matter in determining the price of weapons. CONCLUSIONS This paper has quantitatively investigated the nature of the small arms market. With the benefit of newly compiled cross-country time-series data on the price of AK-47 assault rifles it has been possible to generate empirical findings on previously hypothesized aspects of the small arms market. The model developed to characterize the small arms market is theorized to be driven by four factors - income, motivation, regulation, and supply costs. Estimation of the reduced form version of the model finds that regulation and supply costs are significant determinants of weapon price. This result is robust to various proxies for the concepts. The effective height of 16 trade barriers for weapons, both within and between countries is consistently significant in weapon price determination. Surprisingly, when controlling for other factors, the collapse of the Soviet Union does not have as large an impact on weapon prices as is generally believed. The significance of neighborhood effects, as proxied by neighbors' military expenditure and an Africa dummy (as a residual measure of border porousness) indicates that regional trade is at least as important as global weapons trade. On the demand side, there is some evidence that, for a given level government effectiveness, increasing income raises the price of weapons as a wealth mark-up for a partially non-tradable good. Proxies for the motivation to acquire weapons: lagged income growth, homicide rate, and share of young men do not perform as well as expected. This may suggest that the historic focus on the supply side is justified. More likely, however, it indicates that better modeling and operationalization of the preferences for purchasing weapons is required. A further qualification to the demand side results is that the price data collected are predominantly for the AK-47. By focusing on the AK-47, the most basic assault rifle, substitution effects are ignored if buyers substitute into other higher-grade weapon types as income rises. Further Research The burgeoning field of small arms research has produced a sizeable quantity of survey work. Compiling this growing wealth of survey information into a format amenable to statistical analysis has the potential to provide insights in addition to those garnered from close investigation of single cases. As the first statistical analysis of small arms, this study has uncovered many new empirical questions to consider and illuminated numerous avenues for future research. Data collection This study has begun the task of systematically collecting weapon price data and is intended to be an ongoing project. It is envisioned that the small arms research community will allocate 17 responsibility for collecting statistically useful data in the areas of weapon flows, stockpiles, ammunition price, and border porousness. Collecting these data will be necessary in order to make further quantitative approaches to small arms research possible. Empirical Analysis Cross-country, time-series data on weapon prices will also facilitate the testing of hypotheses on the relationship between small arms and civil conflict. For example, does the availability of small arms (as proxied by price) affect the probability of civil war onset? Does it lead to longer war? Does it result in higher conflict intensity in terms of battle deaths? Investigation of the role of weapons in civil war would seek to evaluate their differential impact on probability of conflict onset, conflict intensity, conflict duration, and post-conflict legacy. Empirical answers to these and other questions will be of direct relevance in generating constructive policy recommendations in relation to small arms policies and managing post-conflict societies. 18 References Boutwell, J and Klare, M. (1999) Light Weapons and Civil Conflict, London: Rowman and Littlefield. Brauer, J (2007) Arms Industries, Arms Trade, and Developing Countries. In Handbook of Defense Economics, Vol. 2, edited by K. Hartley and T. Sandler, Elsevier. Brauer, J and Muggah, R. (2006) Completing the Circle: Building a Theory of Small Arms Demand. Contemporary Security Policy 27 (1) 138-154 CIA (2005) The World Factbook. Washington: United States Central Intelligence Agency. Collier, P and Hoeffler, A. (2004) Greed and Grievance in Civil War. Oxford Economic Papers 56 (4) 563­595. Fajnzylber, P., Lederman, D. and Loayza, N (2002) What causes violent crime? European Economic Review 46 (2) 1323­1357. Kaufmann, D., Kraay, A. and Mastruzzi, M (2005) Governance Matters IV: Governance Indicators for 1996-2004. World Bank Policy Research Working Paper. Kopel, D., Gallant, J. and Eisen, J. (2004) "Global Deaths from Firearms: Searching for Plau- sible Estimates," Texas Review of Law & Politics 8 (1), 113­141. Lacina, B. and Gleditsch, N. (2005) Monitoring Trends in Global Combat: A New Dataset of Battle Deaths. European Journal of Population. 21 (2) 145-166. Lumpe, L., ed., (2002) Running Guns: The Global Black Market in Small Arms, London: Zed Books. Marsh, N. (2007) Conflict Specific Capital: The Role of Weapons Acquisition in Civil War. International Studies Perspectives. 8 (1), forthcoming. Miguel, E., Satyanath, S. and Sergenti E (2004) Economic Shocks and Civil Conflict: An Instrumental Variables Approach. Journal of Political Economy 112 (4) 725­753. Norwegian Initiative on Small Arms Transfers (NISAT) Blackmarket Archive on Small Arms http://www.nisat.org (Accessed September 2005 - January 2006). Small Arms Survey (2004) Small Arms Survey 2004: Rights at Risk. Oxford University Press. Small Arms Survey (2005) Small Arms Survey 2005: Weapons at War. Oxford University Press. Varian, H. (1992) Microeconomic Analysis 3rd ed., W.W. Norton Wooldridge, J.M. (2002) Econometric Analysis of Cross Section and Panel Data. MIT Press. 19 Appendix A: Data Collection Methodology In order to maintain consistency, the exact variable of interest is "the quoted or transacted price in $US for a non-government entity to take possession of an AK-47 assault rifle." Data were sought for four five-year periods from 1986 to 2005. Each price observation is coded with the following details: · Price ($US) · Country · Time period (1986-1990, 1991-1995, 1996-2000, 2001-2005) · The exact assault rifle type observed (e.g. AK-47, AK-74, craft replica) · The location where the price was quoted: (1) city, (2) province or (3) border · Whether the weapon was: (1) new, (2) used, or (3) in need of repair · The source of the price observation (e.g. URL link, reference to published document, name and/or affiliation of field worker) 20 Table I: Descriptive statistics for Kalashnikov prices 1986-2005 Descriptive statistics: 1986-2005 Region Min Max Average Std Dev Observations Asia 40 6000 631 810 81 Africa and Middle East 12 3000 267 417 106 Eastern Europe and fmr Soviet States 50 3000 574 808 75 Americas 25 2400 442 437 59 Western Europe 225 1500 990 443 12 Total Observations 335 Total unique countries 117 Table II: Global average Kalashnikov price Average Price - All Countries Year Ending 1990 1995 2000 2005 All countries 448 425 559 534 Observations per period 46 82 106 101 Table III: Variables for Estimating Weapon Price Determinants Model Variable Observed Variables Weapon price (P) AK-47 assault rifle price Income (I) Per capita GDP (PPP $US) Motivation (M) Lagged per capita GDP growth Civil war onset Young men share Underlying homicide rate Regulation (R) Government effectiveness Democratic accountability Law and order African continent Supply cost (S) Neighbors' military expenditure Own military expenditure Civil war legacy Post-Soviet collapse Distance from Moscow 21 Table IVa: Results of Weapon Price Regression 1 2 3 4 5 6 7 8 GDP per capita PPP 2000$ 0.003 0.004 0.004 0.01 0.01 0.01 0.01 [0.01] [0.01] [0.01] [0.01] [0.01] [0.01]* [0.01]* Neighbours' Military Expenditure -36.55 -29.71 -30.24 -31.87 -29.55 -28.32 -27.28 -31.75 [12.35]*** [12.54]** [10.81]*** [10.93]*** [9.01]*** [10.89]*** [12.98]** [13.55]** Government Effectiveness 215.83 176.17 173.12 135.59 173.4 [59.62]*** [61.89]*** [60.67]*** [56.08]** [60.66]*** GDP per capita Growth, t-1 0.25 0.74 [2.86] [2.97] Civil War Legacy -0.03 -0.02 -0.03 -0.03 -0.03 -0.05 -0.05 [0.02]* [0.01]* [0.01]** [0.01]** [0.01]** [0.02]* [0.02]* Africa Dummy -292.5 -293.87 -394.04 -356.95 -302.34 -332.79 -364.85 [122.54]** [120.93]**[120.78]***[113.85]*** [121.06]** [136.46]**[139.41]*** Ln Distance from Moscow 124.05 125.45 129.76 112.53 125.16 134.17 130.08 [62.66]** [61.54]** [64.20]** [53.57]** [61.52]** [68.80]* [71.45]* Law and Order 2.98 [25.88] Democratic Accountability 33.9 [19.32]* Post-Soviet collapse period -41.42 [30.15] Observations 222 212 228 228 265 228 187 187 Number of countries 85 81 81 81 94 81 69 69 R2 0.08 0.18 0.17 0.18 0.10 0.17 0.18 0.11 Table IVb: Results of Weapon Price Regression 9 10 11 12 13 14 15 16 GDP per capita PPP 2000$ 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 [0.01] [0.01] [0.01] [0.01]* [0.01] [0.01] [0.01] [0.01] Neighbours' Military Expenditure -33.16 -30.1 -34.78 -35.25 -35.46 -32.04 -32.16 -32.16 [12.56]*** [12.45]** [11.37]*** [10.96]*** [10.97]*** [10.90]*** [10.91]*** [10.91]*** Government Effectiveness 175.05 [66.64]*** Civil War Legacy -0.03 -0.03 -0.03 -0.03 -0.03 -0.03 -0.03 -0.03 [0.02]* [0.02] [0.01]** [0.01]** [0.01]** [0.01]** [0.01]** [0.01]** Africa Dummy -325.54 -378.19 -390.35 -331.24 -337.8 -337.8 [126.03]***[126.94]***[125.89]*** [126.47]***[126.95]***[126.95]*** Ln Distance from Moscow 132.02 125.18 90.47 70.81 72.3 120.79 123.66 123.66 [62.89]** [67.57]* [72.46] [64.19] [64.50] [64.06]* [64.27]* [64.27]* Gov Effectiveness 33rd-66th percentile 224.1 -105.61 125.22 [120.90]* [130.42] [120.88] Gov Effectiveness 66th-100th percentile 307.48 230.83 [134.81]** [131.67]* Gov Effectiveness 33rd percentile -257.91 -169.57 -230.83 [107.03]** [107.58] [131.67]* Young Men 26.66 [32.06] War Start 1.5 [70.47] Military Expenditure 14.55 [9.40] Observations 201 196 215 228 228 228 228 228 Number of countries 77 78 76 81 81 81 81 81 R2 0.13 0.17 0.18 0.07 0.08 0.08 0.13 0.14 Standard errors in brackets All regressions contain a constant * = significant at 10%, ** = significant at 5%, *** = significant at 1% 22 Figure 1: Regional Kalashnikov Prices 1200 1000 800 )tnerru 600 (c D US 400 200 0 1990 1995 2000 2005 Year End Asia Africa and Mid.East East.Eu & Fmr Sov states Americas Western Europe Figure 2: Average AK-47 Prices $800 $700 $600 $500 D $400 US $300 $200 $100 $0 1990 1995 2000 2005 Year Countries - civil war Countries - no civil war African countries 23