You are on page 1of 19

This article was downloaded by: [University of Sussex Library

]
On: 15 December 2014, At: 07:36
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Defence and Peace Economics
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/gdpe20

Military Expenditure, Economic Growth
and Heterogeneity
a

b

J. Paul Dunne & Nan Tian
a

School of Economics and SALDRU, University of Cape Town, Cape
Town, South Africa
b

School of Economics and DPRU, University of Cape Town, Cape
Town, South Africa
Published online: 09 Dec 2013.

Click for updates
To cite this article: J. Paul Dunne & Nan Tian (2015) Military Expenditure, Economic Growth and
Heterogeneity, Defence and Peace Economics, 26:1, 15-31, DOI: 10.1080/10242694.2013.848575
To link to this article: http://dx.doi.org/10.1080/10242694.2013.848575

PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &

Downloaded by [University of Sussex Library] at 07:36 15 December 2014 Conditions of access and use can be found at http://www.tandfonline.com/page/termsand-conditions .

2013. There are. natural resources abundance. bSchool of Economics and DPRU. H56 INTRODUCTION There is an impressively large and growing literature on the effects of military spending on economic growth that reflects a continuing lack of consensus. Cape Town. using an exogenous growth model and dynamic panel data methods for 106 countries over the period 1988–2010. http://dx. American University of Cairo.and high-income groups.1080/10242694. Development JEL Codes: O11. across the medium. World military *Corresponding author. Cape Town. School of Economics and SALDRU. The lowest income group continues to experience declines in military spending and military spending as a share of gross domestic product (GDP) remains the lowest for the low-income countries relative to the other income groups (SIPRI. Egypt. The factors considered are different levels of income. E-mail: John. Economic growth. A major focus of the paper is to consider the possibility group heterogeneity and non-linearity. 15–31. which should help researchers in identifying any relation with economic factors.org/10. PAUL DUNNEa* AND NAN TIANb a School of Economics and SALDRU.848575 MILITARY EXPENDITURE. there is more information in the data.ac. South Africa. Having estimated the model for all of the countries in the panel and finding that military burden has a negative effect on growth in the short and long run. 2015 Vol. ECONOMIC GROWTH AND HETEROGENEITY Downloaded by [University of Sussex Library] at 07:36 15 December 2014 J. 2012). and the robustness of the results is evaluated. 26. University of Cape Town. As more data become available that do not reflect the particular geopolitical environment of the cold war. providing strong support for the argument that military spending has adverse effects on growth. some intriguing results that suggest that for certain types of countries military spending has no significant effect on growth. University of Cape Town. 1.Defence and Peace Economics. in final form 12 April 2013) This paper examines the impact of military expenditure on economic growth on a large balanced panel. Keywords: Military expenditure. South Africa (Received 16 December 2012. 21–22 June 2012 and we are grateful to the participants for comments. however. openness and aid. Certainly. University of Cape Town. The estimates for the different groups are remarkably consistent with those for the whole panel. Conflict.Dunne @uct. conflict experience. the end of the cold war saw sizeable reductions in military expenditure internationally and in recent years the declining trend has bottomed out and military expenditure is once again on the increase.za An earlier version of this paper was presented at the 16th Annual International Conference on Economics and Security.doi. the panel is broken down into various groupings based upon a range of potentially relevant factors. © 2013 Taylor & Francis . No.

In most recent empirical work. strategy. interstate). Irrespective of which perspective one takes the topic of military expenditure is. non-trivial. income groups (low. respectively. DUNNE AND N. security and welfare. conflict and income groups. and trade openness. there is no standard framework to fit military spending into. TIAN expenditure in 2010 reached $1. net recipients of aid. MILITARY SPENDING AND GROWTH It is important to have a theoretical model for empirical work. This paper contributes to the existing literature by continuing this investigation with a particular focus on parameter heterogeneity using a post cold war balanced panel of 104 countries for 1988–2010. representing 2. some form of neoclassical growth model has been used. conflict. exogenous or endogenous. 2005). Dunne and Uye (2010) in a survey of 102 studies on the economic effects of military spending show that negative effects of military spending on growth were reported in 39 and 35% of cross-country and case studies.6 trillion. though more recent studies do seem to be providing more . such as the effect of institutions. Smith. while the ‘against’ group sees such spending as a wasteful enterprise that influences the economy beyond the resources it takes up. the answer is almost always an empirical one. however. though Smith did indicate that there could be a small negative effect in the long run. Section 4 presents the results of the estimated growth model using the cross-country data and considers the variation in results across various subsamples when the sample is stratified by developed. and Willenbockel. while over 40% found unclear results. Although the effects of military expenditure have been debated for almost 40 years. A number of researchers have tried to survey the existing literature. non-developed. and Willenbockel (2005). P. The final section presents some conclusions. to provide a consistent and flexible framework for the analysis (Dunne.Downloaded by [University of Sussex Library] at 07:36 15 December 2014 16 J. natural resources. more sophisticated techniques are required to find such a relationship. as there is no agreed theory of growth among economists. Opinions on the effects of military spending are divided among two groups. etc. most definitely. but since most economic theories do not have an explicit role for military spending there is no obvious choice. 1978) which found positive association. civil. which includes military spending and overcomes some of the limitations of earlier models. These will inevitably focus on particular aspects of the growth process and may miss complexities. The empirical debate over the relation between military expenditure and growth was started by the contributions of Benoit (1973. While the end of the cold war meant fewer major international conflicts. Smith. Indeed. Dunne (1996) and Smith (2000) finding no empirical regularity. conflict (conflict or not. they remain a major concern for the developing world. Section 3 provides a discussion of the data-set. The next section briefly reviews the existing literature for developing countries and provides an exposition of the growth model based on Dunne. high). natural resources. often leading to important economic consequences. there was an increase in internal conflicts in the developing world and while the number of conflicts has been declining more recently. positive or negative. Only 20% found positive effects for both types. What is clear is past research has not been able to provided consensus on the economic effects of military spending. Ram (1995). the ‘pro’ group that views military spending as a guarantee of peace.6% of global GDP or equivalent to $236 per person. This led to a large amount of research activity and an impressive build-up of literature that has tended not to support Benoit’s initial findings. medium.

A final issue that has been highlighted in the literature is the impact of openness of economies on growth. Levine. 2013). They argue that in the presence of natural resource abundance. thus making conflict/ rebellion feasible or perhaps even attractive. Natural resource endowment has been investigated as having an impact on conflict and is a good candidate for a factor that might influence the relation between military spending and growth. conflict experience. net recipients of aid and trade openness. natural resource abundance. monetary and trade) and no impact in the presence of poor policies. Collier and Hoeffler (2004) found that where rebellions/civil wars are motivated by greed. The more recent cross-country studies have also shown increasing concern over group heterogeneity. Burnside and Dollar (2000) concluded that aid has a positive impact on growth in developing countries with good policies (fiscal. Tiwari and Shabiz (2012) who considered non-linearities. Looney and McNab (2008) who considered economic freedom and governance. Following on from these contributions. and Dunne (2012) who allows for non-linearity and conflict experience in SSA.Downloaded by [University of Sussex Library] at 07:36 15 December 2014 DEFENCE AND PEACE ECONOMICS 17 consistent support for a negative effect of military spending on growth (Dunne and Tian. but a cumulative battle death of over 1000 throughout the duration of the conflict is also considered. It certainly seems a reasonable to suggest that resource abundance countries may differ in their relation between military spending and growth. There is no consensus. Looting then leads to political instability and hence diminished growth. Smaldone (2006) argued that military burden within Africa generally corresponds to the security realities and affects the relationship between military spending and growth. an unchecked ruler can use resources as collateral and facilitate acquisition of loans and loot the economy. (2011) explains that in a resource-rich country. the concern is with investigating the application of the specified theoretical growth model and secondly many of the factors are invariant over time and so indicators could not be introduced as explanatory variables. natural resource dependence is measured as the ratio of mineral exports to total exports and a country is considered mineral dependent if mineral exports constitute more than 25% of a country’s total exports. endogeneity and non-linearity. Sarr et al. Another potentially important variable is foreign aid. These include Smaldone (2006) who suggested that differences in results for Africa reflect countries’ experience of conflict. developed and developing countries. recent developments have considered the impact of aid on developing countries. In the growth literature. Following conventions in the literature. This sample stratification method is used because firstly. primary commodity exports can substantially increase conflict risk. with no consensus. This is consistent with the IMF’s definition of export dependence. opportunities arise through extortion and looting of profits for those in control of the resource. On the other hand. and Roodman (2004) and Hansen and Tarp (2000) rebut Burnside and Dollar (2000)’s claim and find that aid works for countries with poor policies. Easterly. with proponents of trade openness having a . While there is no consensus regarding the impact of aid on growth. Other research suggests similar hypotheses. Collier and Hoeffler (2004) identify diaspora and their impact on conflict through flows of funds that can support insurrection. it is recognised that aid fungibility can result in discretionary spending. In the conflict literature. Conflict experience would seem to be an important potential source of heterogeneity. As is common in the literature conflict is defined here as having at least 25 combat-related deaths per year. As mentioned. this paper estimates a growth model for the sample of 104 countries and then stratified into sub-samples to compare the results across income groups. Thus. it is reasonable to suppose that the impact of military spending may differ between countries that are net aid recipients and those that are non-recipients. Dunne (2012) provides support to this claim.

There is also a substantial literature on trade and conflict. L is labour and A is the technology parameter evolving according to: AðtÞ ¼ A0 egt mðtÞh (2) where g is the exogenous rate of Harrod-neutral technical progress and m is the share of military expenditure in total output (GDP). The steady-state level of ke is: . The starting point of the model is an aggregate neoclassical Cobb–Douglas production function featuring Harrod-neutral technological progress. Smith. Frankel and Romer (1999) and Dollar and Kraay (2004) and those finding that find trade openness has a negative effect on growth including Yanikkaya (2003). The key assumption within this exogenous growth model is that military spending share m = M/Y affects factor productivity via a level effect on the efficiency parameter. according to SIPRI. whereby the effect of military spending on economic growth is based on the augmented Solow growth model with Harrod-neutral technical progress. 2007). This represents an opportunity cost since the resources used for arms imports could be used for developmental purposes. the top seven suppliers or arms in the world (USA. Russia. ¼ seða1Þ ln ke  ðn þ g þ dÞ @t (3) K where ke ¼ AL denotes the effective capital-labour ratio and α is the constant capital-output elasticity. Putting it differently. The majority of the world’s economies are arms importers. and Willenbockel (2005) is used. K is the real capital stock. P.Downloaded by [University of Sussex Library] at 07:36 15 December 2014 18 J. involvement in the arms trade can impact upon economic growth for a given level of military expenditure (Yakovlev. Loayza. one can display the dynamics of physical capital accumulation by: @ ln ke k_ e ðtÞ ¼ sK kea ðtÞ  ½n þ g þ dk_ e . the UK and Italy) are also within the top 10 of the world’s top military spenders. with a country considered open if it is above the value for the world average and vice versa for a closed economy. France. In 2011. suggesting that countries that trade are less likely to fight (Polachek. and Villanueva (1996). TIAN positive effect on development including Edwards (1998). MODELLING MILITARY SPENDING AND GROWTH For the empirical analysis. DUNNE AND N. but has the potential to have a permanent effect on per capita income along the steady-state growth path. a more open economy could represent greater net arms imports than the equivalent closed economy. with Rodriguez and Rodrik (2000) finding little evidence that open trade policies are associated with economic growth. The share of military spending can also affect the transitory growth rates along the path to the new steady-state equilibrium. Y ðtÞ ¼ KðtÞa ½AðtÞLðtÞ1a (1) where Y denotes aggregate real income. similar to Knight. the model developed by Dunne. Together with the standard Solow model assumptions of an exogenous savings rate s. hence. constant labour force growth rate n and constant rate of depreciation δ. 2007). The openness variable is calculated by taking the sum of a country’s imports and exports and dividing that by its GDP. Germany. In addition. a permanent change in m does not affect the long-run steady-state growth rate. China. which controls Harrod-neutral technical change.

t . Following Dunne. Smith. ye is related to observable per capita income y = Y/L via: n ln yðtÞ ¼ ez ln yðt  1Þ þ ð1  ez Þ  ln A0 þ o a ½ln s  lnðn þ g þ dÞ 1a þ h ln mðtÞ  ez h ln mðt  1Þ þ ðt  ðt  1Þez Þg ð9Þ where z is still equivalent to (α − 1) (n + g + δ). bj ln xj.05 follows from Mankiw et al. and Willenbockel (2005). 2. 2.i. N . u ¼ ln ke . Smith. z ¼ ða  1Þðn þ g þ dÞ (8) Now using Equations (2). T where x1 ¼ s = the gross investment/GDP.t þ g þ dÞ = labour force growth rate plus ðg þ dÞ which is a constant. and Willenbockel (2005) suggest that the dynamic panel model specification can now be written in the form: ln yi. .t . while θ is the elasticity of steady-state income with respect to the long-run military expenditure share. . the linearised form is f ðu Þ þ f ðu Þ½uðtÞ  u  The assumption that (g + δ) = 0..052. .i.t1 þ 3 X j¼1 ln zk.t which is equal to 1 2  0   Re-writing (3) in the form du dt ¼ f ðuÞ.. (7) and (8).DEFENCE AND PEACE ECONOMICS ~k  ¼ e  s nþgþd 19 1=ð1aÞ (4) where the asterisk denotes the steady-state value of the variable. (1992) .t þ 2 X ak (10) k¼1 i ¼ 1.t1 þ gt þ li þ ti. ¼ c ln yi. It is now possible to solve for the steady-state value of output. x3 ¼ mi. t ¼ 1. x2 =ðni. . Equation (9) provides the basis for the empirical analysis Dunne. Linearising (3) via a truncated Taylor series expansion around the steady-state1 level and substituting (4) we get: @ ln ke ¼ ða  1Þðn þ g þ dÞ½ln ke ðtÞ  ln ke  @t (5) Downloaded by [University of Sussex Library] at 07:36 15 December 2014 Y  and since ln ye ¼ ln AL ¼ a ln ke then: @ ln ye ¼ ða  1Þðn þ g þ dÞ½ln ye ðtÞ  ln ye  @t (6) where the steady-state level of output per effective worker is: ~ye ¼  s nþgþd a=ð1aÞ (7) Equation (6) estimates the transitory dynamics of output per effective worker in the neighbourhood of the steady-state level. . assumed to be equal to 0. (6) is integrated forward from t − 1 to t giving: ln ye ðtÞ ¼ ez ln yðt  1Þ þ ð1  ez Þ ln ye . .

DUNNE AND N. We also follow from Knight. while high-middle-income countries were defined as middle-income and the high-income definition left unchanged. (comnat). lt represents group specific effects and vi. and high-income. using data from Haglund (2011) and UNCTADstat. The variable nt reflects the time specific effects. the lowincome and low-middle-income countries were combined to form low-income countries. ( fuel ). TIAN military spending as a share of GDP. As the net recipients of aid is measured as a share of GDP any country that on average received less than 0. A country is given a numerical value of one if it has experienced a conflict between the periods 1988–2010. Other economic variables.t1 which is the lagged variable of military spending and gross investment as a share of GDP.01% of aid as a share of GDP was considered as non-aid recipients.3 In order to homogenise the sample size for the different income groupings. Aid recipients were then divided 3 Country classifications for developed/non-developed are taken from the World Bank. lower-middle income.20 J. Aid was taken from the World Banks WDI. Due to difficulties in obtaining reliable data for the average growth rate of the labour force. with each variable given a value of one or zero. respectively. Income groups are also taken from the World Bank. $1005 or less. while taking g and δ to be constant and exogenous across time and country. dividing countries into those that had experienced either a civil war or interstate war. upper-middle income. . The mineral dependence variable (where mineral exports constitute at least 25% of total exports) was based only on the year 2010 data. gross-fixed capital formation as a share of GDP (used as a measurement of capital) were taken from the World Bank’s World Development Indicators database (WDI). The focus was on the six types of fuels and non-fuel minerals as defined by the Standard International Trade Classification (SITC) codes shown in Table I. indicates whether a country is fuel dependent and the third variable. the second variable. by dividing the sample into countries that are net recipients of aid compared to those that are not.t1 and z2 ¼ si. A civil war and interstate war variable was constructed. Classifications of countries that are either developed or non-developed and income groupings are also taken from the WDI. Downloaded by [University of Sussex Library] at 07:36 15 December 2014 DATA DESCRIPTION The balanced panel is composed of annual data for 104 countries covering the period 1988–2010. real per capita GDP (growth). The groups are: low-income. (non-fuel). records countries that are non-fuel. P.276 or more. $12. characterises whether a country is natural resource dependent via a combination of fuel and non-fuel minerals. and Villanueva (1996) and Islam (1995) in treating s and n as variant across countries and time. The natural resource indicator was divided into three variables.t is the error term. The conflict indicator was taken from the Uppsala Conflict Data Programme and International Peace Research Institute Oslo (UCDP/PRIO) database. mineral dependent. Military expenditure as a share of GDP was taken from the Stockholm International Peace Research Institute (SPIRI). Loayza. $3976–12. the detailed criteria are as follows: the WB divides economies into income groups according to 2010 gross national income (GNI) per capita. $1006–3975. The first variable. z1 ¼ mi.275. but this seems reasonable given the variable is unlikely to show much variation over time. This includes all countries that had a maximum of three missing observations for the military expenditure variable. population growth is used. Natural resource abundance was measured by mineral exports as share of total exports.

0 = non-dependent) Combined fuel and non-fuel natural resource dependence. 4 = Europe.73 6.718 3. America. (1 = dependent. 3 = Asia & Oceania.02 0.49 0.41 0. .15 0.20 0. Real per capita GDP Military expenditure as share of GDP Gross fixed capital formation as share of GDP Natural log of real per capita GDP Natural log of military expenditure as share of GDP Natural log of gross fixed capital formation as share of GDP Natural log of total population Lagged natural log of real per capita GDP Lagged Natural log of military expenditure as share of GDP Lagged Natural log of gross fixed capital formation as share of GDP Growth rate of real per capital GDP (log) Growth rate of military expenditure as share of GDP (log) Growth rate of fixed capital formation as share of GDP (log) Population growth rate (clpop) + 0. 2 = Medium aid. 0 = Non-net recipient of aid) Aid Indicator (0 = Non-net recipient of aid. 3 = High aid) Fuel dependent countries (1 = fuel dependent.30 0. lubricants and related materials. 0 = No conflict) Civil war indicator (1 = Civil war.74 21.39 0.30 0. 0 = non-dependent) Aggregate natural resource dependence (Inc.35 1. 5 = Middle East Income indicator (1 = Low income. other than those of division 56. fuel and non-fuel) (1 = dependent.29 0.39 0.71 3.32 0. 0 = non-dependent) Non-fuel mineral dependent countries (1 = dependent.73 0. non-monetary SITC 3: Mineral fuels (including natural gas).47 0.71 0.02 −0.157 2.25 8. 0 = non-dependent) Openness indicator (1 = open.28 0.30 1.57 1.18 0.02 16.00 −2.43 0. Non-fuel Minerals Fuel TABLE II Variable Description and Summary Statistics Variable name y m k ly lm lk lpop ly1 lm1 lk1 cly clm clk lngdpop dev cont inc.74 12.11 0. petroleum and precious metals) SITC 28: Metalliferous ores and metal scrap SITC 68: Non-ferrous metals SITC 667: Pearls and semi-precious stones SITC 971: Gold. 0 = not open) in 2000 Openness indicator (1 = open.75 0. 0 = No interstate war) Aid Indicator (1 = Net recipient of aid. 1 = Low aid.71 1.17 0.05 (assumed value for g+d) used in Solowstyle regressions Development indicator (1 = Developed. Dev.43 Refer to Footnote 8 and 10 for the difference between the variable nat and comnat.DEFENCE AND PEACE ECONOMICS 21 TABLE I Types of Minerals as Classified by SITC Codes SITC code and description Downloaded by [University of Sussex Library] at 07:36 15 December 2014 SITC 27: Crude fertilizers. 3 = High income) Conflict indicator (1 = Conflict. 0 = Non-developed) Continent indicator (1 = Africa.05 0. 0 = No civil war) Interstate war indicator (1 = Interstate war.84 0.02 0.63 1.31 0. 2 = N & S.73 0.47 0. 0 = not open) in 2009 12.72 3.28 8.76 0.37 0.38 0.75 0. and crude minerals (excluding coal.18 0.75 0.48 1.44 1.93 0.27 2. conflict civilwar intwar aid aid2 fuel nonfuel nat comnata open00 open09 a Variable description Mean Std.35 0. 2 = Middle income.48 0.

00199*** (9.0443*** (−3.0925*** (−7.0273*** (5. we have chosen two arbitrary years in our time-series data.0233** (2.17) −0. further into countries that receive low (less than 1% of GDP). t-ratios in parentheses. 5 A list of countries featured in the sample can be found in Table A1 in the appendix.05.0001 (0.01.28) −4.54) −0.53) −0.0181*** (−3.81) 0.1.0542*** (-6. 76 developing countries. P. ** p < 0.69) −0.63) −0. namely 2000 and 2009 to identify for changes in openness.22 J. medium (between 1 and 3% of GDP) and high amounts of aid (greater than 3% of GDP).16) 0.6) 0.0441 (−0. 20 Asian and Oceania countries.29) 0.53) −0.75) −0.52) 0.02) −0. export and GDP figures are recorded in constant US dollars. * p < 0.0940*** (−10. During the 10-year period countries such as Uganda and China moved from closed to open while others such as France and Greece moved the opposite direction. The sample includes 28 developed countries.0283*** (-4. we were unable to take the sample average of a 23-year period or the initial period (1988).59) 0.0170*** (−4.0431*** (5. 29 African countries.03) −0.430*** (−8. DUNNE AND N. The imports.375 1541 76 0.369*** (−8.22) −0.00933 (−1. 21 North and South American and 10 Middle East countries.5 4 The full sample is divided into 79 open and 25 closed countries for the year 2000 and 78 open and 26 closed countries for the year 2009. TIAN TABLE III The Growth Effects of Military Expenditures in the Solow Growth Model Sample variables clk clm lngdpop Downloaded by [University of Sussex Library] at 07:36 15 December 2014 ly1 lk1 lm1 Year Constant Observations Number of id R-squared (1) All countries cly (2) Developed cly (3) Non-developed cly 0.13) −0. The trade openness variable (trade) takes on the value of one for open or zero for closed economies. .01) −3. 26 European countries.213*** (15.09) −0. Thus. *** p < 0.4 Table II above provides a summary of the final data set containing 104 countries over a 23-year period. Due to missing observations.0554*** (8.97) 2148 104 0.0917*** (−12.00252*** (9.5) 0.131 Notes: Dependent variable: Growth rate of real per capita GDP (cly).08) 0. the GDP figures have been deflated using purchasing power parity.0280*** (−5.0451*** (−4.28) −0.127 607 28 0.0179*** (−4.0230*** (4.

00304*** (6. *** p < 0.00124 (0.76) −0.56) 0.81) −0.0326*** (−3. * p < 0.0935*** (−7.13) −0.0384*** (4.127 0. random effects and random coefficient estimators have been increasingly used and as longer time-series data becomes available dynamic specifications have been introduced into panel data methods (Smith and Dunne.0199 (−0.6) 821 40 0. consistent as T goes to infinity and where T is large the bias will be relatively small.00256*** (6.94) −4.72) 0. There is further heterogeneity bias when the parameters differ over the groups.0214** (−2.120*** (9. ** p < 0. a major problem has been poor data quality and the lack of exogenous variation within the data.0140* (1. and Smith. It is also not consistent as N (number of groups) goes to infinity for fixed T. Stratifying for Income Sample variables clk clm lngdpop ly1 lk1 lm1 Year Constant Observations Number of id R-squared (1) Low cly (2) Middle cly (3) High cly 0. as in the following example of a simple bivariate dynamic model: yit ¼ ai þ b xjt1 þ kyjt1 þ uit (11) A fixed effects estimator would suffer from lagged dependent variable (yit − 1) bias. This has raised a number of issues.96) 695 33 0. 2002).42) −0.77) −0.0820*** (−6.82) −5.92) −0.500*** (−6.76) −0. Smith.00421 (−0.1.19) −0. t-ratios in parentheses.83) 632 31 0.09) −0. It is. but this can be dealt with by estimating TABLE IV The Growth Effects of Military Expenditure.36) 0. which biases the OLS estimator (coefficient of xit − 1) downwards. .0247*** (−3. Nikolaidou.26) 0.67) −0. however.0200*** (−3.64) −0. 2002.000588 (1.03) 0. and Willenbockel. 2004). since the end of the cold war.51) −0.39) 0. Panel data methods such as simple fixed effects.271 0.609 (−0. data quality and leverage has improved and the developments of panel data techniques has helped overcome limited exogenous variation in the data (Dunne.24) −0.01.0265*** (−4.0831*** (−8.DEFENCE AND PEACE ECONOMICS 23 EMPIRICAL ANALYSIS Downloaded by [University of Sussex Library] at 07:36 15 December 2014 In undertaking the empirical analysis of military spending and growth.137*** (11. Dunne.0953*** (−7.366*** (−6.05.0257* (−1.257 Notes: Dependent variable: Growth rate of real per capita GDP (cly). However.0210** (2.

i ¼ 1.41) −0.t1 þ 3 X bj j¼1 ln xj.00320*** (9. t-ratios in parentheses. .33) 0. ¼ c ln yi. Stratifying for Conflict (1) Conflict cly (2) No Conflict cly (3) Civil War cly (4) Interstate cly 0.184 227 11 0.327 Sample variables clk clm Downloaded by [University of Sussex Library] at 07:36 15 December 2014 lngdpop ly1 lk1 lm1 Year Constant Observations Number of id R-squared 773 38 0. t ¼ 1. . x3 is the labour force growth rate or (n + g + δ).24 J.t1 þ gt þ li þ ti.0314*** (3.0715*** (−5. where all variables are in . The estimated general first-order dynamic model takes the form of: ln yi.0227*** (3.t .12) 0.27) 0.0234*** (−3.13) −0.0646*** (7.106*** (−7.58) −5.110 Notes: Dependent variable: Growth rate of real per capita GDP (cly).0320*** (−4. 2.00318*** (8.1. .0193*** (−3.74) −0. N .607*** (−9.0421*** (4. a dynamic model is specified and fixed effects are used to estimate it. P. the overall biases are likely to offset each other due to the estimates of β (downwards) and λ (upwards) working in opposite directions (Dunne.677*** (−3. The re-parameterised general first-order dynamic model is then estimated and the results are presented in Table III.71) −0.34) 0.t .0138** (−2.00264*** (4.51) 674 33 0.61) −0. x1 is gross investment/GDP. 2.28) −4.01.99) −0. 2002).0180*** (−3.84) −0. x2 is military spending/GDP. the resultant heterogeneity bias will bias the estimates of λ upwards.5) −0. In the case of positive serial correlation in the independent variables.00104*** (3.0710*** (−7.0414*** (4. T ð12Þ k¼1 where y is GDP per capita.28) −0. and Smith.85) 0.i.0180* (−1.195 1375 66 0.126*** (−5.81) 0.0765*** (3. * p < 0.52) −5.69) −0..37) 0.0354*** (−2.644*** (−8. .29) 0.0331*** (−4.0281** (−2. each equation individually and taking an average of the individual estimates. Nikolaidou.t þ 2 X ak ln zk.118*** (−9.758*** (−4.43) −1.7) −0.37) −0.123*** (5.86) −0.37) 0.06) 0.0208*** (−4.5) 0. . For the long-run estimates.144*** (−3. . DUNNE AND N. As the data available are not long enough to use large-N and large-T methods.0299*** (3.05.i.02) −0.23) −0. ** p < 0.. TIAN TABLE V The Growth Effects of Military Expenditure.5) −0. *** p < 0.

69) −11.0251*** (2. The empirical growth model is generally well specified across the groups.00366*** (4. For all three income groups. The 5 countries are Bahrain.09) 0.00281 (0. Table III also provides estimation results for developed and non-developed countries.84) 0.793*** (−7.05. Significance levels: *** p < 0. Stratifying for Conflict and Income Sample variables clk clm lngdpop Downloaded by [University of Sussex Library] at 07:36 15 December 2014 ly1 lk1 lm1 Year Constant Observations Number of id R-squared (1) Conflict and low Inc.0575*** (−3. Oman and Saudi Arabia. but only evident in the low.0282*** (−4.106*** (−7.113*** (−3. In non-developed country groups. the effect of military burden on growth is negative and significant in the short run.0302** (−2.181*** (6. representing the change in per capita GDP. there is a negative and significant relation between military expenditure and economic growth in the short and long run.83) 486 24 0. ** p < 0. cly (2) Conflict and med Inc.84) −0. .574 Notes: Dependent variable: Growth rate of real per capita GDP (cly). the full sample was stratified into three different income groups.214 189 9 0. cly (3) Conflict and high Inc. giving the results in Table IV.00775*** (6.37) 0.93) −0. Considering possible heterogeneity in the sample. Kuwait.35) 0.47) 0.63) −0.45) −0.37) −0. giving results which are similar to the full sample.61) 0. all the variables are statistically significant with signs as expected and a clear negative relationship (short and long run) between military expenditure and economic growth. c represents the change in the variable and the dependent variable is (cly). the effect of military expenditure may well be very different for countries with different income levels.1.00323*** (7.67) −5.0875*** (3.0210* (1.78) −0.68) −0.and high-income countries in the long run.146*** (−4.0343*** (−3.0923*** (3.0343* (1.23*** (−5.00729 (−0. Hungary.4) 0. As Dunne (2012) and Pieroni (2007) argue.0226* (−1.356 98 5 0. To consider such differences. suggesting a non-linear relation.41) 0. t-ratios in parentheses. logs. These results show a very well-defined empirical model.12) −0.6 6 The difference in sample size of 5 countries between the high-income and developed countries are due to countries classified as high income in terms of per capita GDP (oil economies) but not developed.450*** (−9.DEFENCE AND PEACE ECONOMICS 25 TABLE VI The Growth Effects of Military Expenditure.01. while for the developed countries there is a negative and significant short-run effect but no long-run effect. cly 0.7) 0.120*** (−4.06) −0.11) −0. * p < 0.11) −6.0705** (−2.73) −0.

58) −0. Peru and United Kingdom. DUNNE AND N.46) −0. conflict and no-conflict groups.165 380 19 0.16) 0.65) −0.94) 0.31) −0.323*** (−8.22) −0. Breaking the full sample into groups of countries that have experienced conflict (38) and those that have not (66) gave the results in Table V. Significance levels: *** p < 0.46) −0. t-ratios in parentheses.26 J.0196*** (−4.6) −0.and high-income groups the results show be analysed with caution.27) −0. particularly in Africa.260*** (−6.01.0202 (1. have found differences in the military spending growth relation for countries in conflict and those not (Dunne. with six countries experiencing both.0750*** (−5.49) −0.49) 0.44) −1. were grouped into low-income. P. Due to the lack of observations for the medium. Iran.65) −0.0945*** (−4. medium-income and high-income countries.12) −5.48) −0. with significant negative short.163 1284 61 0.12) −0. The consistency in the sign and statistical significance of the Solow growth variables even when the sample is limited to 98 observations or five countries (conflict and high income) provides further evidence towards the welldefined nature of our empirical model.0416*** (4.117*** (−5.00783 (−1.657*** (−3. Conflict could have a different effect on countries at different income levels.122 355 18 0.05. which show significant negative effects of military expenditure both in the short and long run irrespective of whether a country has experienced conflict. Previous studies.0309*** (3.8 While 7 Ethiopia.00100*** (3.06) −0.156*** (−4.36) 0.0254** (−2.33) 0.66) 0.24) −6. 8 .192 Notes: Dependent variable: Growth rate of real per capita GDP (cly).12) −0.91) 864 43 0.27) 0. * p < 0. Pakistan.0244*** (3.1.0767*** (8. This gave 33 countries experiencing civil conflict and 11 interstate conflict.96) −0.0507*** (3.00246*** (4.15) 0.149*** (−7. 2012).0327*** (2.87) −0.52) 0.0156*** (−2.0398*** (−3.0472*** (−4. so to investigate the 38 countries that have experienced conflict. India.00612 (−0.0294*** (−4. Stratifying for Natural Resource Sample variables clk clm lngdpop Downloaded by [University of Sussex Library] at 07:36 15 December 2014 ly1 lk1 lm1 Year Constant Observations Number of id R-squared (1) Natural resource cly (2) No resource cly (3) Fuel cly (4) Non-fuel cly 0.0247*** (−3. TIAN TABLE VII The Growth Effects of Military Expenditure.and long-run effects of military burden.7 The estimation results for countries with civil war experiences (Column 3.126*** (−9.0673*** (−7.00365*** (7.07) 0.75) 0. * *p < 0. but the results here do not support this.0240* (−1. Table V) were consistent with the overall sample.55) −4.00310*** (9.0274** (2. It is possible that the type of conflict may be more relevant so Table V also reports results for countries that experienced civil conflicts and those that experienced interstate conflicts.

Interestingly.6) 0.DEFENCE AND PEACE ECONOMICS 27 TABLE VIII The Growth Effects of Military Expenditure.23) −0.76) −1. * p < 0.0468*** (5.10 9 It should be noted that within some natural resource abundant countries (Table VII.45) −0. but Column 3 is consistent with Column 1.and long-run effects. the coefficient (on clm) is more than three times larger than that for the conflict affected low-income countries and larger than the overall conflict and civil war sample and the long-run coefficient (on lm1) is twice as large as the low-income conflict affected countries and more than three times greater than the overall conflict and civil war group. but.37) 0.500*** (−4.00642 (0.01.85) −0.02) −0.0868*** (−5. This suggests that military spending can have a particularly damaging impact on growth in high-income countries involved in conflict (Table VI).0254*** (−4. ** p < 0.108*** (−10.00607 (0.00839 (−0.87) −0.27) 0.331 239 12 0.36) 0.000968*** (2.24) −0.607** (−2.0251** (−2.77) 0.676*** (−8.55) 1325 65 0.213*** (−7.0692*** (−6.43) 0.9 In the rationale for stratifying for natural resource abundance an indirect hypothesis was that resource-rich countries spend on average more on military than non-resource-rich countries.52) 0.01) −0.07) 0. 10 See page 6.0171*** (2.245 390 19 0. Column 1 shows negative and significant short.16) −0.105*** (−5.and long-run effects of similar size for both countries with and without natural resource abundance.05.00306*** (5.0171** (−2.00256*** (4.86) 0.75) −0.0305*** (−4. the combination of all mineral exports exceeds the 25% threshold amount.09) −4. under natural resources.67) 0.0326** (−2.212*** (−6. Stratifying for Net Recipients of Aid Sample Variables clk clm lngdpop Downloaded by [University of Sussex Library] at 07:36 15 December 2014 ly1 lk1 lm1 Year Constant Observations Number of id R-squared (1) Aid cly (2) No aid cly (3) Low aid cly (4) Medium aid cly (5) High aid cly 0.32) 0.165*** (13.51) 0. Table VII shows that military burden has negative and significant short.103*** (−7.32) −4.2) 0.6) −0.82) −0.0376*** (−4. Column 1) one type of mineral export does not constitute more than 25% of total exports. t-ratios in parentheses.87) −5.0377*** (−2.0384* (1.00295*** (6.1.227 696 34 0.35) −0.0427*** (−3.0358** (2.22) −0.131 823 39 0.412*** (−4. .16) −0.22) −4.00270*** (9.151*** (10.41) 0.0182*** (−4.0300*** (3.0212*** (−3.0908*** (−5.92) −0.69) −0.97) 0. the UNCTADstat database was used to divide the sample into 43 countries that are resource dependent and 61 countries that are not resource dependent. To consider the possible impact of natural resource differences.142 Notes: Dependent variable: Growth rate of real per capita GDP (cly).88) −0. Column 2 shows insignificant effects for medium-income countries experiencing conflict.0102 (1.59) −0.41) 0. Significance levels: *** p < 0.0260*** (2.00617 (0.

the negative effect is less for the resource abundant countries.0454*** (−4.16) 0.56) −2.15) −0.3) 0.105*** (−11. This could mean that resources abundance makes military burden more affordable. Interestingly. Bulgaria.0301*** (−5.87) −0.57) 1630 79 0.17) −0.0146* (−1.00207*** (6.115 518 25 0.23) −3. India and Norway.00156*** (5. the coefficient estimates suggest that even though on average natural resource abundant countries spend more on military then non-natural resource abundant countries. Brazil.00198*** (7.571*** (−7.0181** (−2.28 J.84) −0.171*** (−6.0206*** (3. Significance levels: *** p < 0.315 1610 78 0.345 Notes: Dependent variable: Growth rate of real per capita GDP (cly).19) −3.13) −0. Stratifying for Trade Openness Sample variables clk clm Downloaded by [University of Sussex Library] at 07:36 15 December 2014 lngdpop ly1 lk1 lm1 Year Constant Observations Number of id R-squared (1) Open 2000 cly (2) Closed 2000 cly (3) Open 2009 cly (4) Closed 2009 cly 0.74) −4.01.82) −0.0129** (−2.0653*** (−5.0358*** (4. Estimates of the means showed that resource-rich countries allocate on average 3.0310*** (−5.42) −0. Canada.33) 0. .27) −0. but the non-fuel results showing military spending to have no effect on growth in the long or short run.47) 0.0522*** (5.0823*** (−9. DUNNE AND N.0198*** (−4. Albania.44) 0.24) −0. 11 Note: The difference in sample size between the sum of the fuel and nonfuel minerals countries and the natural resource abundant countries are the countries that fit into the profile identified in footnote 12 above.47) 0.162*** (12.0184*** (3. The six countries are namely.0632*** (−4. ** p < 0. This would suggest that the oil economies in the natural resource sample are driving the negative effects of military burden in overall estimates of natural resource abundant countries.72) −0.0611*** (5. t-ratios in parentheses.0477*** (−3.11 with the results for the fuel resource-rich countries consistent with the general results.05. To consider if there was any difference in the type of resources.5) −0.5) 0.103 538 26 0.120*** (−7.17) 0.1.0675*** (−5.4% of their GDP to military expenditure. while non-resource-rich countries spend only 2.0302*** (4.1 percentage points is statistically significant.3% of their GDP on the military.0112** (−2. These show clear differences.73) −0.607*** (−5.00265*** (7. the group was broken down into those that are resource rich in fuel and those rich in non-fuel minerals.906*** (−6.0189*** −4.07) −0.48) 0. P.69) −0. TIAN TABLE IX The Growth Effects of Military Expenditure.3 0.98) −0. giving the results in Columns 3 and 4. * p < 0.9) −0. This difference of 1.154*** (11.26) 0.

. of the 104 country sample only nine countries were classified as net arms exporters. CONCLUSION Military spending by governments is indeed important in the influence it has. albeit only at the 10% level in the short run for closed economies in year 2000. especially when it leads or facilitates conflict. Finally.and high-aid recipients. The results show negative significant short. The results showed negative and significant effects for military spending on growth in both the short and long run. though further stratification by income led to insignificant results for medium-income countries that had experienced conflict. Furthermore. It does seem as though post 12 Stratification of arms imports and exporters was considered as an alternative to using openness. Consistent results were observed when countries were broken into three income groups. thus providing insufficient observations for meaningful regression analysis. Similarly consistent results were found when grouping countries by their natural resource endowments. The mean estimates for military burden for the open economies for the years 2000 and 2009 were 2.12 As the degree of openness could change both the opening and closing value of exports plus imports as a share of GDP was used as a check for consistency. with the short-run coefficient estimate negative and significant for all three groups and the long-run coefficient estimate negative and significant for the low-income and the high-income groups.8 and 3. Smith. the impact of openness is considered. Despite these differences.and long-run effect of military burden irrespective of whether a country receives foreign aid. While military spending on growth for conflict or non-conflict affected countries was both negative and significant. When countries were grouped as developed and developing only the long-run effect for the developed countries was insignificant. since one might expect the effect of military spending to be different for a country that is a net exporter compared to a net importer. aid dependence and trade openness. using the modelling framework suggested by Dunne. As a result the economic impact of such spending is of great concern. Table VIII shows the results for countries that receive aid and those that do not and the stratifies those that receive aid into low-. These results do seem to provide valuable robustness checks and support strongly the view that military spending has an adverse effect on growth. which were about one half a per cent of GDP greater than the closed economies. When the type of conflict was considered. This also holds for countries receiving medium and high levels of aid. but the results shown in Table IX were very similar. both groups showed significant adverse effects of military burden on growth.Downloaded by [University of Sussex Library] at 07:36 15 December 2014 DEFENCE AND PEACE ECONOMICS 29 Moving to consider the possible impact of aid. The estimation results using the dynamic first order model with fixed effects provided surprisingly strong support for the negative impact of military burden on growth for both the short and long run. and Willenbockel (2005) and considering possible sample heterogeneities and non-linearities. medium. However. This paper develops and analyses a comprehensive post-cold war balanced panel dataset for the period 1988–2010. the negative coefficients of the short-run (clm) and long-run (lm1) military burden increases in size as we move higher up the levels of aid received.0% respectively. both civil and interstate wars groups showed negative and significant effect of military spending on growth. the expected differences in the effects of military burden on growth in civil and interstate wars was not apparent. but not for those receiving low levels of aid.

IMF Staff Papers 43 1–44. MA: MIT Press for NBER. R. (1978) Growth and defense in developing countries. Discussion Papers 0708. R. and Uye. In The Global Arms Trade. R. SIPRI (Stockholm International Peace Research Institute).P. The American Economic Review 94(3) 774–780. edited by H..uk/ Armsproduction/Papers/ DirkPaulRonpap3. C. R. A.P. In Handbook of Defense Economics. J. (1973) Defense and Growth in Developing Countries. Tan. development and conflict. Economics of Peace and Security Journal Economists for Peace and Security (UK) 2(2) 60–68. (2000) Aid. D. Amsterdam: Elsevier. (2004) Theoretical and econometric issues in analysing the military expenditure-growth nexus. and Willenbockel. policies. Available at http://carecon.. (1995) Defense expenditure and economic growth. (2004) Aid. Defence and Peace Economics 23(6) 549–557. and Tarp. Defence and Peace Economics 16(6) 449–461. (1992) A contribution to the empirics of economic growth. D. Ram. (1995) Growth empirics: A panel data approach. The American Economic Review 90(4) 847–868. Easterly. P. edited by B. Feder. In Industrialization and Growth: A Comparative Study. W.. (1998) Openness. R. 261–338.. Smith. (2008) Can economic liberalization and improved governance alter the defensegrowth trade off? Review of Financial Economics 17 (3) 172–182. and Roodman. D. Dunne. In Macroeconomics Annual 2000. Rogoff. S. growth and poverty. Dunne. Syrquin.. and Hoeffler. (2004) Trade. D. H. Hartley and T. Sandler. Loayza.P. (2012) Military spending. certainly no evidence of any significant positive effects. care is needed in applying the conclusion to individual countries. Economic Development and Cultural Change 34(2) 261–275. Feder. (1996) Economic effects of military expenditure in developing countries: A survey. policies and growth. (2013) Military expenditure and economic growth: A survey. however. (1999) Does trade cause growth? The American Economic Review 89(3) 379–399. Knight. growth. Gleditsch. J. R. D. and Weil. 293–305.N. and Tian. C. edited by A. . Haglund. and growth: Comment. Romer. Polachek. P. Bernanke and K. E. London: Routledge. June) How trade affects international interactions. investment and economic growth in small industrialising economies. N. Smith. and Smith. E. J. (1983) On exports and economic growth. and Rodrik. G. Dunne. as some groupings of countries did differ. (2010) Military spending and development..P. Levine. Mankiw. There is. University of the West of England. D. N. Pieroni. J.and Middle-income Countries. M. R. low-aid recipient countries showed no effect of military spending on growth. J. and Kraay. Chenery. E. and Dollar.. (2000) Trade policy and economic growth: A skeptic’s guide to the cross-national evidence. Benoit. CA: Regina Books. (2004) Greed and grievance in civil war. The Quarterly Journal of Economics 107(2) 407–437. TIAN cold war it is possible to see this result developing as a consensus view. T. E. G. S.M.pdf. (2000) Aid effectiveness disputed.org. Sarr. Robinson and M. Downloaded by [University of Sussex Library] at 07:36 15 December 2014 References Benoit. P. Rodriguez. The Economics of Peace and Security Journal 8 (1) 5–11. D. E. and Swanson. and McNab. Burnside. MA: Heath. 439–464. Oxford: Oxford University Press.J. Dunne.. D. South African Journal of Economics 70(5) 789–790. Dollar.A. Bulte. Oxford Economic Papers 56(4) 563–595. M. (2011) Blessing or Curse? The Rise of Mineral Dependence Among Low. 251–273. S. J. Public Choice 148 353–380. (2007) Military Spending and Economic Growth. productivity and growth: What do we really know? Economic Journal 108 383– 389. for example non-fuel natural resource abundance. Collier. New York: Oxford University Press. Edwards. edited by N. (1996) The peace dividend: Military spending cuts and economic growth. DUNNE AND N.30 J.N. M. Journal of International Development 12(3) 375–398. A. (Various years) Yearbooks World Armament and Disarmament. Hansen. In Making Peace Pay. Quarterly Journal of Economics 110(4) 1127–1170. D. Lexington Books. Oxford: Oxford Policy Management. (1986) Growth in semi-industrial countries: A statistical analysis.P. Dunne. But while a compelling conclusion can be drawn across countries. (2005) Models of military expenditure and growth: A critical review. Mimegraph. Dunne. 263–282. G. Islam. Department of Economics.P. (2007. Nikolaidou. P. J. N. Frankel.. Dunne.W. (2011) On the looting of nations. and Romer. and Willenbockel. Cambridge. and Villanueva. Claremont. Economic Journal 114(493) 22–49. Looney. Boston. (2002) Military spending. edited by K.P. D. D. Journal of Development Economics 12 59–73. Meissner. F.

R.R. M. Yanikkaya. CA: Regina Books. de Soysa. Smith. 15–24. and Dunne.P. (2012) Does defence spending stimulate economic growth in India? A revisit. Tiwari. Gleditsch. (2000) Defence expenditure and economic growth. (2003) Trade openness and economic growth: A cross-country empirical investigation. P. Defence and Peace Economics 18(4) 317–338. and economic growth. R. Smith. Claremont. Yakovlev. P.P. Lindgren. G. Birkbeck College: Mimeo. N. and Shahbaz. In Making Peace Pay: A Bibliography on Disarmament and Conversion. H. Fiji India Indonesia Japan S. military spending. (2002) Military Expenditure Growth and Investment. J.K. Smit and I. Korea Malaysia Mongolia Nepal New Zealand Pakistan Papua New Guinea Philippines Singapore Sri Lanka Thailand Europe Albania Austria Belgium Bulgaria Cyprus Denmark Finland France Germany Greece Hungary Ireland Italy Luxembourg Malta Netherlands Norway Poland Portugal Romania Russia Spain Sweden Switzerland Turkey UK Middle East Bahrain Egypt Iran Israel Jordan Kuwait Lebanon Oman Saudi Arabia Syria . Mouhleb. (2007) Arms trade. Journal of Development Economics 72 57–89. APPENDIX TABLE A1 List of Countries in Full Sample Africa Algeria Angola Botswana Burkina Faso Burundi Cameroon Djibouti Ethiopia Ghana Kenya Lesotho Madagascar Malawi Mali Mauritania Mauritius Morocco Mozambique Namibia Rwanda Senegal Seychelles Sierre Leone South Africa Swaziland Tansania Tunisia Uganda N & S. edited by N. Ecuador El Salvador Guatemala Jamaica Mexico Nicaragua Panama Paraguay Peru United States Uruguay Venezuela Asia & Oceania Australia Bangladesh Brunei Cambodia China. J. Defence and Peace Economics 1–25. (2006) African military spending: Defense versus development? African Security Review 15(4) 18–32.P. A. S.DEFENCE AND PEACE ECONOMICS 31 Downloaded by [University of Sussex Library] at 07:36 15 December 2014 Smaldone. America Argentina Belize Bolivia Brazil Canada Chile Colombia Dominican Rep.