Julien Malizard

According to SIPRI, world defence spending is equal to $1.739 billion in 2017, which represents nearly 2% of the world GDP. Given its economic importance, the field of defence has been investigated by many economists. Defence economics relies on the use of all the economic approaches relative to the defence sector. The first articles published in the field mainly focused on industry (microeconomics), alliance (public economics), arms transfers (international economics) or economic impact (macroeconomics). More recently, the scope has been enlarged to new topics such as civil wars, terrorism or new areas, such as China.

From a methodological point of view, a major difficulty arises when it comes to measure defence output. Indeed defence spending is just a sum of inputs (wages and equipment) but does not give any figure of the efficiency of the allocation process (which is the core of economics). The key issue of “how much is enough” remains involved. As a consequence, defence economic academic literature says little about the main purposes of defence and focus only on the pure economic effects.

This chapter reviews the main methods used by economists in order to evaluate the causes and consequences of the defence sector: narrative approach, theoretical model and econometric analysis. Each method is discussed in length to examine their weakness and forces.

Introduction

According  to the latest figures published by the think tank, Stockholm International Peace Research Institute (SIPRI), world defence spending (DS) was USD 1,750 billion in 2018, which is the highest level in history, surpassing the levels previously observed during the Cold War. On average, the share of DS is close to 2% and seems rather small compared to previous periods (Sandler and George  2016). There are microeconomic  figures corre- sponding to these macroeconomic  figures and, in particular, those related to the equip- ment market. From SIPRI figures on the defence industry, the total turnover of the top 100 is close to USD 450 billion. Flows of trade are difficult to measure but a recent esti- mation indicates that arms exports are less than USD 100 billion.

Given  the economic  importance  of the defence  sector,  economic  evaluation  of defence  matters a great deal. Of course,  the  main  goal of defence  is to provide security, and economic factors should play a limited role. However,  given the impor- tance  of DS  at the  macroeconomic   and microeconomic   levels,  the  relationship between economics  and strategy is of interest.1  Economic  literature provides useful information  on  many  aspects of defence  policy, such  as  the  armament  industry, defence budget or economic impact of defence activities. The aim of this chapter is to discuss the economic  approach  of defence  studies and to review  briefly  the main themes of the academic literature in economics.

Following  Robbins, “Economics is the science which  studies human behavior  as a relationship between ends and scarce means which have alternative uses” (1935, p. 16). Defence  economics should not be different  than economics:  an economic  approach requires a comparison of ends and means. However,  the key issue is to convert military ends and means into economic ends and means. Strategic means are obviously  difficult to evaluate and, thus, economists often rely on defence policy in order to investigate military ends. Military  means are more easily converted into economic means because there is a price tag for each item.

The relationship between military ends and means has been investigated by Malizard(2019). He stresses the importance  of “strategic autonomy” for the French defence policy  that provides the framework  for strategic  ends. Strategic  autonomy  is barely defined, but refers to nuclear deterrence, the choice to conduct (alone or in a coalition) a large spectrum of military operations, and the necessity to rely on French equipment; the French interests include not only its national  territory,  but also overseas depart- ments. Given  these ends, France armed forces require a large spectrum of equipment. The military  means (in terms  of soldiers  and equipment)  are then converted  into economic means: a large defence budget compared to other European countries, and an independent defence and industrial base (DIB). These economic  means are more easily analysed by economists as they refer to the budget (macroeconomics)  or industry (microeconomics),  which constitute the two main areas in economics.

Standard economic evaluation is based on “cost-benefit approach” (CBA). For each alternative, CBA requires data on what is paid (cost) and what is gained (benefit). The net result provides an economic rationale for each alternative and one may rank all the alternatives to get the best solution, which is called “an optimum”  in economics. In macroeconomics,  such a method is difficult to apply because there is no data on each alternative; in such a case, the solution  is to directly analyse the effect of one variable on another, possibly by directly comparing the two alternatives.

Unfortunately, CBA cannot be used for defence economics. The main reason is that the benefits are not measurable. Consider a rise in DS by USD 1. How will this rise affect security? Does this dollar translate into something more than, less than, or equal to one dollar in terms of security? These questions remain unsolved  as security does not have (monetary) proxies.

This chapter is about the main economic approaches applied to the defence sector. By doing so, I follow closely the Intrilligator’s definition (1990, p. 3):

Defence  economics is concerned with that part of the overall economy involving defence-related  issues, including  the level of defence spending, both in total and as a fraction  of the  overall economy;  the  impact  of defence  expenditure,  both domestically for output and employment and internationally for impacts on other nations; the reasons for the existence and size of the defence sector; the relation of defence spending to technological  change; and the implications of defence spend- ing and the defence sector for international stability and instability.

As Smith (2009, p. 2) points out “Defence and peace economics, like much of the rest of economics,  uses a mathematical  and statistical language  that makes it incompre- hensive to many who might be interested but do not have a quantitative training”. This chapter is written in a manner so that it can be fully comprehended by all researchers, whatever their level of competence in mathematics.

Economic  reasoning is well established in, at least, what is called “mainstream eco- nomics”:  markets  provide a useful framework  to investigate  economic  behaviour because this is the best way to achieve “optimality”, except in certain  cases, such as externality, asymmetry of information, or public goods. Economic modelling relies on theoretical models that use the concept of markets; then, the model is confronted with data. One issue is to know what should come first between data and model. Data are not neutral per se and reflect theoretical  considerations. This point is worth remem- bering as many economists do not wonder how data are built. With these remarks in mind, I first discuss economic  modelling  applied to the defence sector and, second, data related to defence economics.

The chapter is organized  as follows.  The first section provides an insight on “how much  is enough”, which,  as already  discussed, is the key issue for economists. The second section presents the main theoretical approaches and the third section the main empirical analysis. As the scope of economics  is vast, a selection of the most investi- gated topics in defence economics are presented in the fourth section. The final section concludes this chapter with a discussion of the future of defence economics.

The main issue: how much is enough?

A lot of actions  can be analysed through  economic  thinking.  The decision  to do something is related to its costs and benefits. As long as the costs are smaller than the benefits, the action is called profitable.  Difficulties arise when the researcher must con- sider all the direct and indirect  costs and benefits and translate them into a single measure, such as a monetary one.

Markets are among one of the most useful concepts used by economists. Markets help in determining both the supply and demand for a single good. The interaction of supply and demand may lead to an equilibrium, with an “optimal”  quantity of the good supplied and bought at a certain price. Markets allow economists to measure the monetary  relationships between two agents. Defence  industry and labour forces con- stitute the supply side of the market as they provide  inputs for defence. Governments represent the demand side of the market as they provide security.

Economists evaluate complex consequences. These consequences can be either posi- tive or negative and one has to take into account all these elements in an economic evaluation. Some consequences are easier to capture than others. In the case of defence, the consequences are mostly non-economic. In particular, the primary aim in defence is to ensure security. However,  the links between defence, which can be measured from a macroeconomic perspective with a defence budget, and security cannot be measured precisely. Following  Smith (2009), defence activities rely on inputs,  such as soldiers, equipment, experience, or trainings. Defence inputs2 are “bought” by governments by utilizing  their defence  budget.  These  inputs are known,  but  the  defence  output  is unknown.  One cannot  measure,  in economic  terms, defence  capabilities  except  by estimating the inputs required for creating the said capability.  Security can, thus, be considered the sum of all the capabilities needed, given a defence policy that is suppo- sedly representative of citizens’ preferences. From this point of view, security is the sum of goods and services provided by the “public planner” expressed in monetary terms.

In the  case  of  defence,  defence  production  remains  unknown.  For example, consider an increase of EUR 1 of the defence budget in France. What  are the con- sequences for defence output? Are  the inhabitants of France safer by EUR 1? For economists,  there is no simple relationship between the amount of money  spent on defence and security gained. If security is accurately provided,  one empirical possi- bility is to measure the economic gain by comparing it to a hypothetical situation of conflict. This is difficult, given the methodological  assumptions required to compute a situation that does not exist.

If security is not measurable, one may investigate the benefits of defence. According to Hartley (2012),  the  benefits  of security  are both economic  and non-economic. Defence  economics is mainly interested in economic  benefits of defence, such as the defence sector’s impact. Non-economic  benefits are difficult to measure and, therefore, rarely evaluated.

The problem of measuring the benefits has a straightforward, but rather unsatisfy- ing solution.  Economists  consider output to be equal  to the  sum of inputs.  This solution is also applied  to public spending in other sectors, particularly health and education; in these sectors, there is a “private” counterpart, whose effectiveness can be computed and compared with that of public spending. In defence, there are few private counterparts; these are mainly private military companies that are relevant for specific defence missions.

Finally, note that the problem implies flaws not only in macroeconomics, but also in microeconomics. On the macroeconomic side, the key issue is to investigate whether the size of the defence sector is appropriate,  with respect to both budgetary and strategic constraints. On the microeconomic  side, one major problem is to identify defence firms, which is quite difficult in the case of dual-use firms. More specifically, their inputs are potentially developed for military products, civil products, or both.

Building theoretical models

Economics  relies on theoretical models to simplify  social interactions between eco- nomic actors. The main goal of the models is to disentangle the main economic chan- nels of interaction  between these actors and, then, evaluate the best public policy,  if required. Thus, economists have to identify the actors, how they are involved with each other, and the result of their interactions. By comparing the model’s equilibrium with an optimal situation, one may gauge the pertinence of public interventions.

Markets are among  the main formal  approaches to discuss economic  interactions and the way to reach an equilibrium. Markets aggregate both supply and demand and the interactions between generate an equilibrium of both price and quantity. If this equilibrium  has undesirable properties, such as idle capacity, public policy intervention is needed.

Even if the defence sector is quite special, markets offer an interesting perspective for both macroeconomics  and microeconomics.  For instance, the  demand  for defence arises from government and is measured by DS; the supply side of defence is composed of all the sectors that provide inputs.

This section discusses how defence economists  analyse the defence sector in terms of theoretical  models.  I assume that  markets  are useful  in synthesizing  this literature because they offer the possibility of discussing some general economic  features. The concept of markets can be “aggregated” to analyse both the microeconomic  and mac- roeconomic aspects.

Microeconomics

Microeconomics  is the branch of economics  that studies the interactions  of indivi- duals  and firms  within markets  by choosing  the  optimal composition  of scare resources that maximize utility (individuals) or profit (firms). It is widely  accepted among economists that optimality  and desirable outcomes may be reached through markets, but sometimes market failures arise, and then, public authorities must intervene to change individual behaviour.

Conceptually, individual behaviours are determined by a maximization problem. An economic agent seeks to maximize its utility or profit under budget or costs constraints. Optimal behaviour is determined thanks to this problem and leads to a situation where marginal cost is equal to marginal benefit.

Several assumptions (“perfect competition”) are made,  such as symmetry  of infor- mation or multiplicity of actors (on both demand and supply sides), so that they do not influence the optimal price. Within this framework, markets are the best way to achieve optimal outcomes. In this context, prices reflect incentives. However,  there are several cases of market failure, such as externalities, public goods or asymmetry of informa- tion, that require public intervention.

DS is considered  a public good. Because of its features  of non-rivalry and non- excludability,  this is a special type of good in economics. Non-rivalry  implies that the consumption  of this  good by an  economic  agent is  not at the  expense  of the consumption of the same good by some other economic agent. The property of non- excludability   indicates  that  the  consumption  of  this  good does not require  the consumer to pay for it (or no one can be excluded from its consumption). Rivalry  and excludability define private goods. Common goods are a mixture of rivalry and non- excludability and club goods a mixture of non-rivalry and excludability.

Public goods cannot be provided by private players as they have no incentive to enter the market; given both non-excludability and non-rivalry, a firm is not able to make a profit. Public interventions are, then, required because governments may levy  taxes to finance public goods. However,  the features of public goods may also imply free riding, so that some countries in an alliance have an incentive to not provide the optimal level of DS.

The concept of public good implies that defence is provided  collectively.  This may involve the defence being provided directly by the public sector, by the private sector or by a mix of the two (Macdonald  2010). Defence remains a “final” public good, but one that requires a sum of private goods (e.g. military equipment for which there is both rivalry and exclusion) and human capital. From a practical point of view, it is a ques- tion of determining for the Ministry of Defense how much should be achieved “in- house” and which missions can be delegated to private sector companies (outsourcing). Traditionally, the private sector provides equipment, but since the end of the Cold War, many services have been carried out by external suppliers (i.e. clothing, catering). As Bellais et al. (2014) point out, the US and the UK practice outsourcing with a mixed record. Finally, some companies known as “private military companies” participate in defence missions, either alongside or replacing the regular forces. Singer (2001) recalls that  they are used  for operational  activities,  consulting (strategy  and training)  or operational and logistical support. The rise of these companies has led to a blurring of the line between regal and private activities (Markusen  2003) and hence justifies the need to regulate them.

A microeconomics-based  approach is useful to analyse markets related to the provi- sion of equipment or labour. In these markets, one has to identify the economic agents. In the case of equipment, a government  is the buyer and the defence firms are the sellers. As explained by Sandler and Hartley  (1995), this situation leads to a “mono- psony”  where there is only one buyer, but several sellers. Moreover,  economics gen- erally seeks to identify the best alternative from a set of different ones. In the case of defence, in many situations, there is only one choice. As argued by Hartley (2012), lack of incentives is likely to lead to inefficiencies.

Microeconomics theory fails to fully encompass security issues related to defence. Gov- ernments’ goals are linked to not only the economy,  but also strategic preferences. For instance, the equipment market in Europe exhibits low concentration, with national cham- pions enjoying monopoly  in their own country, but harsh competition in other markets. Economic  theory indicates that this situation would require concentration to reduce pro- duction costs through economies of scale. As discussed by Belin et al. (2017), this situation is not likely to occur, given countries’ preferences in terms of equipment. For example, strate- gic autonomy  is the French’s defence policy  pillar; industrial nations, such as Germany, Italy, and the UK, use the defence sector as a way to do industrial policy, whereas other countries prefer American supplies. Moreover,  in terms of use of equipment, individual choices are not identical, as shown by the multiple versions of the aircraft A400M.

Macroeconomics

As stressed in the introduction, the size of the defence sector is significant in terms of macroeconomic  measures (GDP, public spending, and trade, to name a few), but some countries bear a greater “burden” than others. For example, in 2018, Oman spent 8.2% of its GDP on defence; this accounted for 19% of its public spending. Thus, it seems reasonable to question the influence of DS on the economy.

From a macroeconomic point of view, economists evaluate all the channels through which defence affects the global economy. Obviously,  there are many channels – some are direct (such as the multiplier), whereas others are indirect ones. Summing up all the effects of the different channels may provide a measure of defence’s influence. However, quantifying all the channels is difficult  because it requires a vast amount of informa- tion. As a consequence, the economic  literature focuses mainly on one channel. This subsection is based on the surveys by Dunne et al. (2005) and Malizard  (2011). Three main channels have been identified.

The first channel is the “demand channel.” It mainly relates to Keynesian theory. The basic idea is that an increase of DS implies an increase in GDP by a number called the multiplier. The existence and the size of this multiplier  has been widely  discussed by economists since the 2008 economic  crisis; the key issue is to know whether the multiplier is below, equal to, or above 1. As part of public spending, DS may play a role. There is also a negative indirect impact to DS. As discussed by Smith (1980), DS and private investment compete for a fixed share of the savings, so that an increase in DS leads to less resources for investment because of the increase in the interest rate. This situation is called crowding- out and may reduce the final effect of DS from the demand side.

The second channel is the “supply channel”. Rather than focusing on the quantity of DS, as  in the demand  channel, this channel analyses the quality  of DS: does DS increase the quality of inputs? As discussed by Ruttan (2006), American  DS led to several technological  breakthroughs that have been reused in a civilian context. Others, such as Schumpeter (1934), consider that DS is unproductive.  The final impact, then, depends on the quality of the final expense: defence R&D expenditure is more likely to generate  a positive  influence  than defence  compensation.  Ram (1993)  discusses at length  the positive  and negative  supply-side impacts of DS and, in particular,  the debate on technological spillovers and inefficiency of defence industries.

The last channel is called the “security channel”  and relies mainly on the primary purpose of defence: ensuring security. Two opposite effects have been identified. In the first, a country may gain greater security when DS is increased  because rule of law, especially property laws, is enforced. The second one is related to arms race because an increase in a neighbour’s DS can be perceived  as a threat to national  security that requires an increase in DS by the country under threat; consequently, the level of DS can be above the optimal level, leading to major inefficiencies. The final effect is the sum of these two effects, but given the absence of data for measuring security, this issue remains unresolved.

The way these channels are modelled is crucial. Focusing on a specific channel may imply an incomprehensive approach. Depending on the preferences of the researcher, this can lead to biased  results. In this vein, Malizard  (2011) shows that within the inconclusive  macroeconomic  literature,  regularities  in terms  of results arise when papers are classified according to the underlying theoretical model. Modelling  choices are not neutral.

For the purpose  of interpretation, economic  models  tend to distinguish between short- and long-run effects. The former is mainly related to the demand-side channel, whereas the supply-side channel witnesses the latter. The basic idea behind this inter- pretation is the following:  a DS shock may affect the economy in the short run because of extra public spending flowing  into the economy  – this is particularly useful when unemployment prevails. In the long run, the effect is more likely  to arise because of changes in productivity or technological spillovers.

Empirical issues

Having  described the theoretical models in the previous  section,  the next step is to check whether the conclusions of these models are in line with data. Since the end of World War II, national accounts have gained credibility and data are available to the scientific community. Notably,  this allows economists to publish more papers based on empirical analysis (Angrist  et al. 2017) and provide more insightful recommendations on public policy. Moreover,  the development of econometric theory provides reliable tools to identify the causal relationship between two variables.

However, in defence studies, there still exist major gaps in measuring defence activ- ities. For instance, there is no “defence sector” in national accounts. As mentioned in the first section, defence production  is not measurable. Because the defence sector is characterized  by secrecy, many  of the  measures  used  do not provide  an accurate econometric analysis. Data exist but are subject to criticism. This section is about how data are used by defence economists despite the various limitations.

Two types of data are available. From a macroeconomic point of view, the main figures are defence budget and arms trade. From a microeconomic point of view, the key issue is to determine the turnover associated with defence activities in a company.

This  section first discusses some relevant  sources. Then,  I briefly present the main problems  associated with these data: perimeter,  accuracy and econometric  issues due to data limitations.

Data

Economists enjoy providing and using economic data. This is a language which also requires strong academic backgrounds to be comfortable  with. However,  there is little discussion among them on the appropriateness of these data because most papers are “technical”, rather than being theory oriented. As stressed by Smith (1998, p. 423), “many articles  give the  impression  that  the  authors  have  applied  the  statistical procedures without looking at the data”. In this subsection, I review some main data sources.

Macroeconomic data

Quantitative  measures of the defence sector rely mainly on defence budget or the size of the army, both in terms of personnel and equipment, and provide  researchers the opportunity to discuss defence policy.  Two main sources are worth mentioning.  The first, SIPRI, has gained much of its visibility  because of its long-lasting evaluation  of defence budget,  as well as that of arms transfers and defence industry. The Interna- tional Institute for Strategic Studies (IISS) provides useful information  with respect to the size of armed forces and defence budgets. Other sources, such as data published by national authorities in charge of DS, can also be mentioned.

Defining a defence budget, at least theoretically,  seems easy, but there are numerous practical difficulties. One definition  is commonly  used by international organizations and  is based  on the NATO  definition  that includes operating  costs (personnel  and pensions),  procurement,  construction,  and R&D. Statistics  from the  Ministry  of Defense  and other ministries can be used, as long as their budgets are aligned with those of military forces. Even if the definition appears clear, many problems arise when it comes to evaluating  the  budget,  such as inclusion  of paramilitary  forces, direct funding  of the  defence  industry  through  other public  expenditures,  peacekeeping operations, and so on. Malizard  and Richter  (2018) discuss the different sources of these problems  in France and the UK and show major gaps between them that ulti- mately lead to different empirical estimates. It appears that, for France, DS is over- estimated by SIPRI, but the NATO figures suffer from major breaks.

One major drawback of DS is that it is an imperfect measure of power and security. DS does not reflect the military  capabilities of one country and the capabilities are subject to depreciation, which is not considered in official publications. The size of the armed  forces  is, then, a complement  of DS because  it provides  useful information regarding the ratio of troops to equipment and, ultimately, military capacity.

Trade data are used to gauge the intensity of the arms trade between two nations. This kind of data requires information  regarding the quantity of equipment (including all the intangibles, such as training, technological  offsets, etc.), the price, whether the equipment is new or refurbished, and the time between order and delivery. It is crucial to get a single measure that encompasses all the production costs from different coun- tries, with different production  systems, as was the case in USA and USSR  during the Cold War.

SIPRI is the international  standard  in arms trade  data and serves  as a primary source for the World Bank database. The database is rich enough to analyse trade flows between countries since 1950. SIPRI develops its own measures, such as trend indicator value (TIV), to compute trade. TIV does not measure the monetary value of the arms trade; rather, it relies on production  costs for comparable  features of specific arms. Other data sources, such as Arms  Control  and Disarmament  Agency (ACDA)  or national data, are not used in the literature given their lack of transparency and com- parability (Malizard  2018).

Microeconomic data

It is worth remembering that DS has a microeconomic  counterpart because public spend- ing generates economic activities for firms or workers, both in the private and public sec- tors. In the case of defence, microeconomic data rely mainly, on defence firms.

Given  its production volume and turnover, the defence industry is subject to numer- ous debates. These  potential  controversies  require  the most transparent and reliable data to perform  an  accurate empirical  analysis. Two mains  sources regarding  the defence industry have been used in the economic literature. The first one is provided by SIPRI, which cover the top 100 defence firms, and discusses some parameters related to their performances;  these include the turnover,  the share of turnover  because of military activities (which is, in a sense, one measure of the duality of a firm), and the number of employees. This database is easily downloadable  from the SIPRI website and covers the period 2002–2017; older data are available in print in the form of SIPRI Yearbooks. The publications from Defense News are another source.

Although  these data may be useful for quantitative analysis, they suffer from some limitations. First, the sample may be limited. Often, the time period covered is quite short and does not include the period of transition following  the end of the Cold War that shaped the defence industry in terms of its concentration  (Bellais  et al. 2014). Besides, the ranking only includes 100 firms and, thus, neglects smaller firms, which could be potentially  crucial for procurement. Second, the activities of firms are not decomposed into domestic turnover and external turnover; this, in turn, complicates the firm-level  analysis of arms trade. Finally,  for dual-product  firms, the number of employees, R&D, and capital (i.e. the inputs) are not separated into civil and military outputs. This is a frustrating  issue as economists  cannot  discuss the performance  of firms by comparing  inputs and outputs. One solution is to assume that the share of inputs is equal to the share of outputs. This solution is simple to implement but does not fully capture the process of production, especially the potential cross dependencies between civilian and military productions.

Problem with data

For empirical analysis, it is crucial to measure precisely the same economic  phenom- enon. Thus, cross-countries studies require a single measure of defence activities to avoid any bias of over- or under-estimation. Countries do not have the same inter- pretation of DS and,  as already  stressed earlier,  international  institutions  such as SIPRI may lead to a bias. Moreover,  for a single country, one may observe a break in the definition;  for example,  France’s 2009 exclusion of its paramilitary  forces (gen- darmerie)  from the defence  category  has not yet been corrected  by NATO  in its official statistics.

Different  sources are used to compute DS. SIPRI lists three  categories.  Primary sources consist of budgetary documents provided by the government. Secondary sour- ces are quotes from  primary  sources, such as NATO or IMF; other secondary sources are less reliable. According  to Perlo-Freeman  and Ferguson  (2015), primary  sources have increased since 1988 and constitute almost 80% of data in the most recent period; this gives confidence  in the reliability of data. However,  the authors warn the recent increasing trend of absence of data for numerous countries.

Cross-country  comparisons  are  also complicated  due  to the  purchasing  parity conversion. A dollar bill may have different purchasing power in each country and standard market exchange rate does not adequately address this problem. One solu- tion is to consider the purchasing power  parity (PPP) exchange rate. As noted by Smith  (2009, p. 91),  the  difference  on converting  Chinese DS  from the  market exchange rate to the PPP exchange rate is huge – the value of the latter is almost four times that of the former.

The last difficulty with data is comparison over time. Macroeconomic  data are sub- ject to major quantity and price changes over time. Thus, it is necessary to control for the change in price, captured by the GDP deflator, to give real values of the measured economic  variable.  In the  case of defence,  there  is no “military deflator”  and the common  solution consists of using GDP deflator  as a proxy for the military deflator. This solution is unsatisfactory because defence inflation appears to be higher than that indicated by the GDP deflator (Hartley 2016).

Opportunity costs are widely used by economists to discuss all available  alternatives in making an economic decision. In the case of defence, one major issue is the impos- sibility to quantify all the alternatives: we observe DS or wars  but there is no real situation in which there is no DS or the absence of conflict,3  even if one can calculate counterfactuals to compare the alternatives.

Transparency is a key issue in DS. Many  data are subject to classification because they are related to national  security issues. The economic  literature uses open source data when they exist; since the beginning  of the 2010s, SIPRI does not provide any information on many Middle  Eastern countries. Developing  countries may either lack statistical institutions or, given the nature of their political regimes, lack the incentives to provide an accurate measure.

Trade data also suffer from severe limitations. Even if a large number of countries ratify  the arms trade treaty, it seems that data still lack transparency. According  to Malizard (2018), SIPRI data are a kind of “black box” offering few insights on the way data are computed. Further, because SIPRI uses TIV measures, rather than monetary ones, data are not comparable.

Econometrics  is a useful tool as it relies on statistical routines to evaluate the sig- nificance of the relationship between different variables that are based on theoretical models. The development  of new  econometric  approaches allows  the researcher to apply more effective routines by improving  the precision of the estimates. Given  the power of the econometric  methods, one can distinguish between correlation  and caus- ality, except in a case of theoretical misspecification.

Two types of econometric  problems may arise: the consequences of measurement error due to data limitations and reverse causality. According  to Smith (1998), empiri- cal studies are more interested in “using cooking tricks” instead of amending the model to fit data. These tricks potentially  lead researchers to use the best empirical  results in terms of statistical significance given the potential publication bias with “star wars”,4 as stressed by Brodeur et al. (2016).

The data limitations  that have been discussed above  may have potentially  huge consequences on econometric  results. Greene  (2003) shows that the estimates may suffer from a severe bias. There is no good solution to avoid this problem but check- ing the  collected  database  and investigating  sources to ensure that  there  are  no breaks, is one of the obvious  steps in any empirical  analysis. Malizard  and Richter (2018) indicate that different  measures of DS lead to different conclusions about its economic impact.

Econometric  analysis allows the researcher to investigate  the causal influence in a given relationship by distinguishing endogenous (dependent) and exogenous (indepen- dent)  variables. For instance, growth  theory  indicates that GDP is  “explained”  by investment and DS; however, it is as likely  that GDP explains investment or DS, so that the variables considered exogenous variables prove to be endogenous. This is called “reverse causality”. This problem  is very serious in empirical economics and requires specific remedies. One of them  is to consider whether the system formed  by all the variables is endogenous (each variable is explained by all the others), but this solution needs many  observations.  Instrumental variables  have been also used  by econome- tricians; this involves a two-stage model in which one has to find an “instrument” that is related to exogenous variables, but not to the endogenous ones. In practice, finding a suitable instrument is difficult.

The defence economic literature,  a critical survey

The defence economics literature embraces many different subjects incorporating economics tools relevant to the defence sector. I decided to focus on subjects that have been widely examined by numerous authors and correspond with the central features explained above: how do we deal with security issues without  taking  them accurately  into account?

The outline of this section is the following.  First, I discuss the literature on alliance with a specific focus on burden sharing. Second, I analyse arms trading and compare the causes and consequences. Third,  I review the main results of the literature on the defence industry and R&D. Finally, I assess the economic  impact  of DS.

Alliance and burden sharing

Defence  spending  is considered  a public  good. Given  its nature, public  economic theory emphasizes that the government  is the only agent that has the ability to pay for defence. The share of national wealth that is allocated to defence is subject to numer- ous factors (Smith 1989). One such factor is alliance membership.

Olson  and Zeckhauser  (1966)  offered  the first  formal analysis  of alliance  using economic theory. The authors discussed the production  of security from NATO and concluded that it is a pure public good. This good is produced at the expense of richer economies that bear a disproportionate  share of the burden while smaller economics have a tendency to free-ride. Empirically,  the authors show a significant correlation between the share of DS to GDP and gross national product (GNP) for each ally. This result is called the exploitation  hypothesis.

However, after 1970, NATO changed its doctrine, and more recent contributions do not support the conclusion of Olson and Zeckhauser (see, for instance, the survey from Sand- ler and Hartley 2001). The change in doctrine led to different benefits for alliance members (deterrence, but also more local defence activities). Therefore,  the costs of membership alliance are more aligned with the benefits, and the exploitation  hypothesis no longer holds. For many scholars, the key issue now  is to evaluate the degree of publicness of international security provided by an alliance (see Gates and Terasawa 2003).

From an empirical point of view, many researchers have addressed burden sharing within NATO. For instance, Khanna and Sandler (2002) computed a benefit function that includes GDP, population, and area and compare it with cost (which is simply the relative  share of an ally’s DS to the  sum of all individual  defence budgets). Until recently, these papers showed that benefits and costs are in line so that the exploitation hypothesis is rejected. However,  a recent paper by Shimuzu and Sandler (2012) indi- cated that, since 2004, NATO has exhibited divergence between costs and benefits, and the authors conclude that there is a risk that NATO could become a two-tier alliance. The costs-benefits approach is simple and flexible enough to be extended to other alli- ances such as the EU (Kollias 2008).

Nonetheless,  this approach  suffers many  drawbacks.  Costs  are mainly proxied by defence budgets, which do not have qualitative  effects: France spends less than 2% of its GDP on defence,  but its military  capabilities  are probably greater than Estonia, which respects the 2% criteria. Benefits are also subjective and subject to criticism: they mainly  include what a country wants to protect (its population,  its wealth,  and its area), but do not reflect any preference in terms of defence policy such as interventions in military operations or contributions to international organizations.

Trade

Arms  trade  is controversial  among  economists;  there is an obvious ethical concern given  the nature of arms. Arms  trade is related  to the industrial organization  of a country’s defence sector but also to attitudes toward  international security and inter- national treaties. A critical question regarding arms trade is the following:  what are the drivers that explain  the export  and import of arms?  It is important  to distinguish between economic and security factors. Another  subject of debate among economists and social scientists, is related to the use of arms and whether arms trade has a stabi- lizing effect (i.e. deterring aggression) or a destabilizing effect (i.e. fuelling conflict).

Initially, it is appealing  to analyse global  arms trade as a market. One feature of arms exports relies on its extreme concentration. Only  a few countries export arms: according to the SIPRI database, the top 10 countries exported  96% of global arms during the Cold War and 90% after the Cold War. The United States and the Union of Soviet Socialist Republics (USSR) (later, Russia) are the market leaders followed  by some European countries. Emerging  countries are still marginal  (China  ranked sixth after the Cold War), but their influence is growing. Turning to the demand side, arms imports are less concentrated: during the Cold War, the top 10 countries accounted for less than a third of arms imports and, after the Cold War, the top 10 countries’ share was close to 50%. Given  these statistics, standard economic theory indicates that the supply side should have substantial market power. However,  economic factors play a minimal role in the explanation of the global arms trade while political and strategic factors emerge as crucial factors.

The analysis of arms trade  as a standard market  has been a focus of researchers. Anderton  (1995) summarized this literature, which is mainly theoretical. As emphasized by Levine et al. (1994), two opposite drivers may lead to arms exports: economic profit and external security. Arms exports provide insecurity for the seller when the buyer is a rival or security when the buyer is a friend. Arms exports show that depending on the quality of the buyer, the level of exchange is different from a pure economic model. These results explain some features of the global arms trade, but standard economic theory does not fully capture the frictions arising from the peculiarities of arms.

Following new political economic  theory, some researchers analyse arms trade using political  factors such as proxy  of democracy, political orientation, or security factors such as alliances, conflict,  and neighbours engaged in conflict.

Some papers evaluate the decision to export arms (proxied by a dummy that takes the value of one if one export  is observed or zero otherwise) and the quantity of exports. From an econometric point of view, this type of model is known as the Heckman model. Blanton (2005) provided an example of this model and showed that in the selection stage, democracy  is a key positive determinant of US exports; that is, the more democratic the country, the more  likely  the United  States is to export arms to that country. Comola (2012) insisted on the importance of political factors as an explanatory variable for arms exports. More  recently, Martinez-Zarzoro  and Johannsen (2019) show that alongside economic factors, both political and security factors play a significant role based on a sample of 104 exporting countries over the period 1950 to 2007. The authors also note major shifts due to the end of the Cold War. These papers offer interesting insights on arms exports, but they fail to circumvent endogeneity:  the papers use conflicts as expla- natory  variables, but they do control for the fact that trade may have consequences regarding conflicts. This type of circularity could potentially cause reverse causality.

Arms trade generates effects beyond commercial operations because of the nature of arms, their effects in terms of security, and their influence on economic  structures. As acknowledged by SIPRI (1971, p. 73), “perhaps the most important  question about arms supplies is … what effect they have on the development  of wars – on the like- lihood of wars breaking out, on the course of wars and their general severity”. In this vein, several  papers  cast  doubt  on the  destabilization   hypothesis  as  they do not conclude any significant  impacts concerning  arms trade on conflicts (Moore 2012). Fauconnet et al. (2019) indicate that French arms trade does not exhibit any significant role on intrastate conflicts, whereas arms trade in the rest of the world fuels conflicts. Unfortunately,   these papers  cannot  address  the  endogeneity  issue and may imply reverse causality.

Industry and R&D

Among all the studies on defence economics, those that discuss the defence industry are most closely related to microeconomic theory and, in particular, industrial orga- nization.  Standard economic  theory  provides  a useful framework  within which to investigate  the defence  industry, but the specificities  of the defence  sector  such as sovereignty technology,  the importance of fixed costs, and barriers to entry, among others,  must  be  considered.  These  features  imply that  the  equipment  market  is somewhat unique: there are only a few firms and one buyer (the government),  and these entities  form a monopsony. In some  cases, the  context  is called  a bilateral monopoly if there is only one firm.

Defence  industries represent the “supply side of the market for defense equipment” (Sandler and Hartley  1995, p. 177). All firms included in the defence industries define the defence  industrial base (DIB). Defining  the DIB requires a perimeter which is, unfortunately,  difficult  to operationalize.  For instance,  should  housing  services  be included in the perimeter? Dunne (1995, p. 402) had a clear definition by distinguishing three types of products: lethal large  or small  weapons  systems, non-lethal  strategic products (fuel), and other  products consumed by the military  (food). Reliable DIB perimeters are crucial for public policy  as they may shed light on the efficiency of a country in providing  arms. For example, should a country rely on its own capacity? The notions of self-sufficiency and security of supply are crucial determinants in con- structing a DIB because arms can be produced domestically or through international trade with partners. In this vein, the procurement choices of the Ministry of Defence are multiple  and correspond  to a trade-off   between  national  sovereignty,  cost of equipment and existence of industrial capacity. Not all countries have the same strate- gic ambitions and needs, so acquisition policies are diverse (Hartley  and Belin 2019).

As the survey by Hartley  (2007) noted, the literature is mainly composed of papers on R&D issues such as defence financing  and its impact on the economy,  procurement, and industrial cooperation. This subsection briefly follows  this outline.

Many Western  armed  forces rely on technological  superiority for their  missions. Consequently, defence equipment is considered highly advanced in terms of complexity (Mowery  2010). The innovation  concept is complex and widely analysed through the lens of R&D (a measure of inputs) or patents (a measure of outputs). The means of innovation are crucial, and some  papers follow their evolution  (Mowery 2010): the defence sector absorbed a large  share of budgetary  and financial  means during the Cold War, but  these  shares  declined  after 1990. Nonetheless,  for many Western countries, the defence sector is at the core of their national system of innovation given the architectural  knowledge;  that is, the way that knowledge  is applied  to optimize production by assembling parts from different subcontractors.

Relative to innovation output, patents are used by economists, and there is a need to control  the quality (rather than the quantity)  of innovation.  Patents are identified in technological   classes, but defence  is not one of them. Therefore,  the identification of defence patents is complicated and requires industrial expertise. Some indicators mea- sure the influence of a patent, and papers show that the economic  impact is significant due to their effect on productivity (Moretti  et al. 2017).

Data  measurement is the crucial issue for this subfield: patent behaviour  is suppo- sedly the same among  civilian  and military  sectors, which is not the case given  the secrecy associated with the military  sector. For example, Acosta et al. (2017) showed that defence firms patent less than civil firms. Moreover,  the statistical approaches used by economists acknowledge  that a firm with 1% of its turnover accounted for by the military  sector  is a firm of the defence  industrial base. There  is obviously  a limit because given the different defence submarkets and their industrial organization,  the gradient must range between 0% and 100% (and cannot be a binary measure).

The procurement issue is crucial because it allows us to analyse the industrial choices of states in terms of armaments. Given the importance of R&D, the price of weapons systems is rising faster than civilian inflation (Hartley  2016). From  this point of view, it is  necessary  to arbitrate  between  the  cost over the  entire  life cycle (procurement, maintenance, dismantling)  and the need to master technologies crucial to sovereignty (Rogerson 1995). Contractual mechanisms (fixed-price contracts or cost-plus contracts) are used to assess the risk sharing inherent in the production  process (Hartley 2007). In order to avoid cost overruns, fixed-price  contracts are the standard choice in many developed countries (Bellais et al. 2014).

Industrial  cooperation  is a popular  subject in Europe. There are many firms and programmes, and duplication is well-established (Hartley  1995) compared to the inte- grated American  equipment market. Economic  theory shows that cooperation  is eco- nomically  optimal  given  the costs savings (Sandler  and Hartley  1995). However,  as reviewed by Hartley  and Braddon (2014), both the demand and supply sides generate inefficiencies,  particularly  the complexities  that cause partners to collaborate.  These inefficiencies are likely to increase costs and delays.

Economic impacts

A main subject in the defence economics literature is the economic impact of military activities. As DS is a component of public spending, many papers address the influence of DS on macroeconomic   performance  such as GDP, unemployment, or investment. However,   as defence  activities  are also localized  in a few areas  (for historical  or geographical  reasons), another arm of the literature examines the impact  of DS on regional  activities. From this huge  literature  body, little consensus emerges  on the macroeconomic effect of DS. On average, based on a sample of estimates in the litera- ture,  Yesilyurt  and Yesilyurt  (2019)  indicated  that DS has  no significant effect  on growth. DS is not a crucial determinant of growth (Sala-i-Martin  et al. 2004), and its effect is minimal compared to other factors such investment.

There are two explanations for the impact of DS. The first is related to Keynesian theory:  an increase (or decrease) in DS positively (or negatively)  affects economic activities; a multiplier effect. The second  reason  is linked  to the quality  of defence inputs: defence activities are widely technological,  and the technologies conceived and used for the military generate spillovers to the civilian sector. Ruttan (2006) provided many examples of such military technologies: nuclear, semi-conductor etc.

Direct  economic  impacts have received much attention from economists. This  is because the impacts are easier to assess from  an econometric  point  of view. The sim- plest way to model a direct relationship is to use only two variables, DS and GDP, in level form or defence burden and growth rate of GPD in a relative form. This approach has been widely  used because of a lack of global theory encompassing all the channels through which defence affects the economy.  This type of model  is close to the one chosen by Benoit (1972) in his seminal publication. Another  advantage is the fact that the endogeneity  issue can easily be controlled using proper econometric tools. These atheoretical approaches lead to interesting results, notably, the existence of a multiplier effect. However,  there are several significant limitations. The approach neglects oppor- tunity costs: What  would have been the effect if public spending had been included? Finally,  as mentioned by or Dunne and Smith (2010), the results are sensitive to the econometric specification.

Many economists   use  a theoretical  model given  the  limits of the  atheoretical approach. Growth theory provides a useful framework  to investigate the influence of public spending, and the theory has been popularized among defence economists for its solid foundations (Dunne et al. 2005). Many empirical analyses focus on both defence and non-defence spending and compare their respective effect on GDP growth among other  determinants  such as private  investment  or population growth. This approach tends to show a negative impact of DS.

Some indirect models have been used. These models evaluate the existence of a drain of resources that, ultimately, reduce economic performance. Smith (1980) popularized a model where the crowding-out  effect is estimated considering that private investment and public spending compete for the same pool of resources. An increase in DS implies less resources for private investment and a higher interest rate. DS has a strong nega- tive impact on private investment, which is the key determinant for economic growth (Sala-i-Martin 1997). However, DS composition  matters as shown by Malizard (2015) French DS implies crowding-out,  but French defence equipment spending fosters pri- vate investment.

Finally, some models rely on multiple equations that encompass direct effect, indirect impact, and endogeneity of DS (Deger  and Smith 1983). The system of equations  is based on a growth equation to quantify  the direct effect, an investment equation to model the indirect effect, and a DS equation to circumvent endogeneity. Many papers show that DS has a positive  direct  impact  that is surpassed by a negative indirect impact leading to a global negative impact.

As explained by Malizard (2011), some regularities emerge from the global picture of the results. Some models are well-designed  to exhibit a positive effect of DS, whereas others lead to the opposite conclusion. A comparison of models may lead to proper evaluations relative to DS impact. Additionally, the defence economics literature is somewhat original in the global  literature: except for growth estimates, the results are based on specific models with- out points of comparison with other macroeconomic papers.

To summarize the economics of DS, the primary purpose of defence is not to foster economic activity but to ensure security. Economists are unable to discuss the effect of security. If DS exhibits a positive effect, it could be a type of double dividend in a sense that both security and economic  activities are stimulated. However,  as Smith (2009) noted, a high macroeconomic  impact cannot be expected given  the relatively  weak influence of DS in terms of its percentage of GDP.

Conclusion: What is the future of defence economics?

In their introduction  to the second volume  of the Handbook  of Defense Economics, Hartley  and Sandler (2007) acknowledged  that defence economics has changed. From a semantics point of view, a prominent journal in the defence economics field changed its name from  Defense Economics to Defense and Peace  Economics. This change is a recognition  that defence economics is interested in economic  means but also strategic ends such as security, peace, war, and conflicts.

Peace and conflicts economics are not, by far, the core of defence economics from a historical point of view. Conflicts  economics  is mainly  interested in the causes, consequences, and costs of all forms of violence (interstate conflicts, civil wars, and terrorism, among others). The modelling perspective is reversed: defence economics considers conflict  as exogenous  whereas conflict  economics  uses defence as exogen- ous. There  is a kind of simultaneity between defence and conflict. However,  the lit- erature is now  intertwined  as economic  factors are crucial  in the explanation  of conflicts  (Collier and Hoeffler 2007)  and conflicts  have  huge  consequences  on economies (Abadie and Gardeazabal  2003). Besides, the concept of security web has been popularized  by Dunne et al. (2003) by gathering data on both conflict and defence sources.

The  crucial  issue for defence  economics  is  to measure  output:  security  is  a multidimensional  concept  that  cannot  be summarized  in  a single index.  The absence of conflicts  is one item that can be included as a means of defence. From a statistical point of view, a conflict  is coded  as 1 so that 0 has different  meanings (Fauconnet  et al. 2019):  the  durable  absence  of conflict  or a peaceful  period between two periods of conflict.  Statistical measures must be cautiously evaluated as peace  (i.e. 0) is the  absence  of conflict  in this approach.  However,   these data are  used  by many  economists  and social  scientists  to discuss conflict   but  also development issues.

From an economic  perspective,  there is still a missing link: translating  defence means into monetary measures. In my opinion, “How much is enough?” would be a question without answer. Therefore, defence economics is then a complement to other disciplines, notably  strategy and international  relations. However,  it helps  policy- makers to formulate  better informed  description and prescription when economics tools are relevant.

Notes

1  Smith (2009) discusses the notion  of “military value chain” and shows how economics is ori- ginal among social sciences, especially because of the analysis of means in economic terms.

2  See the first subsection of our second part for a discussion on how governments  source its national defence. Note that these inputs are private goods, despite the fact they are collectively procured.

3  Abadie  and Gardeazabal  (2003) investigate the costs of terrorism for the Basque region by

considering a counterfactual region in which there are no terrorist attacks.

4  Statistical significance is labelled in table thanks to “ * ”.

References

Abadie, A., Gardeazabal,  J. (2003). The economic  costs of conflict: A case study of the Basque

Country. American Economic Review, 93(1), 113–132.

Acosta, M., Coronado, D., Ferrandiz, E., Marin, M. R., Moreno P. J. (2017). Patents and dual- use technology:  An empirical  study of the world’s  largest defence  companies.  Defence  and Peace Economics, 29(7), 821–839.

Anderton, C. H. (1995). Economics of arms trade, 523–561. In: Hartley, K., Sandler, T. (Eds.) (1995). Handbook of Defence Economics. Defense in a Globalized World, Amsterdam, Elsevier. Angrist, J., Azoulay,  P., Ellison, G., Hill, R., Lu, S. F. (2017). Economic research evolves: Fields

and styles. American Economic Review, 107(5), 293–297.

Bellais R., Foucault, M., Oudot, J.-M. (2014). Economie de la défense, Paris, La Découverte. Belin  J., Droff, J., Malizard,   J. (2017). The economics of European defence: Some additional

insights, ARES – Comment, 15, available online: https://www.iris-france.org/wp-content/uploa

ds/2017/04/Ares-Group-Comment-avril-2017-.pdf.

Benoit, E. (1972). Defence and Economic Growth in Developing Countries, Lexington,  MA, Lex- ington Books.

Blanton,  S. L. (2005). Foreign policy in transition? Human rights, democracy  and U.S.  arms exports. International Studies Quarterly, 49, 647–667.

Brodeur, A., Lé, M., Sangnier, M., Zylberberg, Y. (2016). Star wars: The empirics strike back.

American Economic Journal: Applied Economics, 8(1), 1–32.

Collier, P., Hoeffler, A. (2004). Greed and grievance in civil war. Oxford Economic Papers, 56,

563–595.

Comola, M. (2012). Democracies, politics and arms supply: A bilateral trade equation. Review of

International Economics, 20(1), 150–163.

Deger,  S., Sen, S. (1983). Military  expenditure, spin-off  and economic development.  Journal of

Development Economics, 13(1–2), 67–83.

Dunne,  J. P. (1995). The defence  industrial base, 399–430. In: Hartley, K., Sandler, T. (Eds.) (1995). Handbook of Defence Economics: Defense in a Globalized World, Amsterdam, Elsevier. Dunne,  J. P., Perlo-Freeman,  S. (2003). The demand for military spending in developing  coun-

tries: A dynamic panel analysis.  Defence and Peace Economics, 14(6), 461–474.

Dunne,  J. P., Smith, R., Willenbockel, D. (2005). Models of military expenditure and growth: A

critical review. Defence and Peace Economics, 16(6), 449–461.

Dunne,  P., Smith,  R. (2010).  Military expenditure  and granger  causality: A critical review.

Defence and Peace Economics, 21(5–6), 427–441.

Fauconnet, C., Malizard,  J., Pietri, A. (2019). French arms exports and intrastate conflicts: An

Empirical investigation. Defence and Peace Economics, 30(2), 176–196.

Gates, W. R., Terasawa, K. L. (2003), Reconsidering publicness in alliance defence expenditures: NATO expansion and burden sharing. Defence and Peace Economics, 14(5), 369–383.

Greene, W. (2003). Econometric Analysis, 5th ed., Upper Saddle River, NJ, Prentice Hall.

Hartley, K. (1995). Industrial policies in the defence sector, 459–490. In: Hartley, K., Sandler, T. (Eds.) (1995). Handbook of Defence Economics: Defense in a Globalized World, Amsterdam, Elsevier.

Hartley, K. (2007).  The arms  industry,  procurement  and industrial  policies,  1139–1176. In: Hartley, K., Sandler, T. (Eds.) (1995). Handbook of Defence Economics. Defense in a Globa- lized World, Amsterdam, Elsevier.

Hartley, K. (2012). Conflict  and defence output: An economic perspective. Revue d’Economie

Politique, 122(2), 171–195.

Hartley, K. (2016). UK defence inflation and cost escalation. Defence and Peace Economics, 27 (2), 184–207.

Hartley, K., Braddon, D. (2014). Collaborative  projects and the number of partner nations.

Defence and Peace Economics, 25(6), 535–548.

Hartley, K., Belin, J. (2019). The Economics of the Global Defence Industry, London,  Routledge. Hartley, K., Sandler, T. (Eds.) (1995). Handbook of Defense Economics: Defense in a Globalized

World, Amsterdam, Elsevier.

Intrilligator, M. D. (1990). On the nature and scope of defence economics. Defence Economics,1 (1), 3–11.

Khanna,  J., Sandler, T. (1996). NATO burden sharing: 1960–1992. Defence and Peace Econom-

ics, 7(2), 115–133.

Kollias,  C. (2008). A preliminary  investigation of the burden  sharing  aspects of a European

Union common defence policy. Defence and Peace Economics, 19(4), 256–293.

Levine, P., Sen, S., Smith, R. (1994). A model of international arms market. Defence and Peace

Economics, 5(1), 1–18.

MacDonald, P. (2010), The Economics of Military  Outsourcing. PhD  thesis, University  of York. Malizard,  J. (2011), Dépenses militaires et croissance économique. Ph.D.  thesis manuscript, Uni-

versity of Montpellier  1.

Malizard,  J. (2015). Does military  expenditure crowd-out  private investment? A disaggregated perspective for the case of France, Economic Modelling, 46(1), 44–52.

Malizard, J. (2018). Taking stock on the measure of arms trade: The case of France, unpublished mimeo. Malizard,  J. (2019). Introduction  to the special issue on France. Defence and Peace Economics,

30(2), 129–132.

Malizard,  J., Richter,  F. (2018). Comparing  defence output: a methodological  nightmare? Con- ference paper presented at the 22nd International  Conference  on Economics and Security at Güzelyurt (Morphou).

Markusen, A. (2003). The case against privatizing national security. Governance, 16(4), 471–501. Martinez-Zarzoso, I., Johannsen, J. (2019). The gravity of arms. Defence and Peace Economics,

30(1), 2–26.

Moore, M. (2012). Selling to both sides: The effects of major conventional weapons transfers on civil war severity and duration. International Interactions, 38(3), 325–347.

Moretti,  E., Steinwender, C., Van Reenen, J. (2014, July). The intellectual spoils of war? Defense R&D, productivity  and spillovers. Conference  paper presented at the 2016 American  Eco- nomic Association Annual Meeting.

Mowery, D. C. (2010). Military R&D and innovation, 1219–1256. In: Hall, B. H., Rosenberg, N. (Eds.) (2010). Handbook of the Economics of Innovation, Vol. 2, Amsterdam, London, Elsevier. Olson, M., Zeckhauser, R. (1966). An economic theory of alliances. Review of Economics  and

Statistics, 48(3), 266–279.

Perlo-Freeman,  S., Ferguson,  N. (2015).  Capacity  and openness: the determinants of the avail- ability and quality of military expenditure data. Conference paper presented at the 19th Inter- national conference on economics and security, Grenoble.

Ram, R. (1993). Conceptual linkages between defence spending and economic growth, 19–39. In: Payne,  J. E., Sahu, A. P. (Eds.) (1993). Defence Spending & Economic Growth, Boulder, CO, Westview Press.

Robbins,  L. (2007  [1935]).  An Essay  on  the  Nature  and  Significance  of Economic  Science, London, Macmillan.

Rogerson, W. P. (1995), Incentive  models  of the  defense  procurement  process, 309–346. In: Hartley, K., Sandler, T. (Eds.) (1995). Handbook of Defense Economics: Defense in a Globa- lized World, Amsterdam, Elsevier.

Ruttan, V. W. (2006). Is War Necessary for Economic Growth? Military  Procurement and Tech- nology Development, Oxford, Oxford University Press.

Sala-i-Martin, X. (1997). I just ran two million regressions. American Economic Review P&P, 87 (2), 178–183.

Sala-i-Martin, X., Doppelhoffer, G., and Miller, R. I. (2004). Determinants of long-term growth: A Bayesian averaging of classical estimated (BACE) approach. American Economic Review, 94 (4), 813–835.

Sandler, T., Hartley, K. (1995). The Economics of Defence, Cambridge, Cambridge University

Press.

Sandler, T., Hartley, K. (2001). Economics of alliances: The lessons for collective action. Journal of Economic Literature, 39, 869–896.

Sandler T., George,  J. (2016). Military  expenditure trends for 1960–2014 and what they reveal.

Global Policy, 7(2), 174–185.

Sandler, T., Shimizu, H. (2012). NATO burden sharing 1999–2010: An altered alliance. Foreign

Policy Analysis, 1–18.

Schumpeter,  J. A. (1934). The Theory of Economic Development: An Inquiry into Profits, Capital, Credit, Interest and the Business Cycle,  Harvard,  CA, Harvard University Press.

Singer, P. W. (2001). Corporate  warriors: The rise of the privatized  military  industry and its ramifications for international security, International Security, 26(3), 186–220.

Stockholm International Peace Research Institute (SIPRI) (1971). The Arms Trade with the Third

World, Solna, Paul Elek.

Smith, R. (1980). Military  expenditure and investment in OECD countries, 1957–1973. Journal of Comparative Economics, 4(1), 19–32.

Smith,  R. (1989).  Models  of military  expenditure.  Journal of Applied  Econometrics,  4(4),

345–359.

Smith, R. (1998). Quantitative  methods in peace research. Journal of Peace Research, 34(4),

419–427.

Smith, R. (2009). Military  Economics, the Interaction of Power and Money, London,  Palgrave

Macmillan.

Yesilyurt, F., Yesilyurt, M. E. (2019). Meta-analysis, military expenditures and growth. Journal of

Peace Research, 56(3), 352–363.