Guns vs butter: Understanding the Impact of Global Defence Spending on Economic Growth in the 21st century.

By Meisha Lukman and Ivan Beales

Introduction

The “guns vs butter debate” is quintessentially an economic question. It deals with the issues of scarcity and resource allocation, describing the trade off faced by governments when choosing to spend on defence or domestic programmes. The phrase was likely first coined over 100 years ago by William Jennings Bryan (Goddard, n.d.), and has become increasingly relevant ever since.

This debate is pertinent today amidst a background of ongoing conflict in almost every corner of the world — the Russo-Ukrainian war, Israeli-Palestinian conflict, civil wars on the African continent, to name a few. As a result, this has led to a massive increase in defence spending across the globe with the Stockholm International Peace Research Institute reporting that global military spending has reached its highest yet at a staggering $2400 billion (SIPRI, 2023), largely fueled by European spending in response to the conflict in Eastern Europe. For context, governments in Central and Western Europe spent around $345 billion in 2022 — surpassing cold war spending (1989). The same can be said for the USA which now spends $877 billion (~3.0% GDP), more than the next 10 countries combined and it has recently passed a fiscal bill approving $866 billion to fund its military (Zurcher, 2023).

This leaves one to question how exactly governments can partake in such spending given that global debt levels are also at an all time high — $307 trillion in 2023 according to the IMF (Masterson and North, 2023). Another pertinent question to discuss would also be the potential economic effects of such military spending, ‘guns’ on economic growth and the production of consumer goods, ‘butter’. Russia, for instance, initially entered the war with a fiscal surplus due to strong cash reserves. However, due to the unexpected length of the war, military expenditures have led Russia into a fiscal deficit (Marrow, 2023). Despite this, in its proposed 2024 budget, defence spending totals 6% of gross GDP, exceeding social spending (Luzin et al, 2023).

In a world which is increasingly insecure (United Nations Development Program, 2022), along with a looming climate change crisis, rise in mental health conditions (i.e., depression, anxiety) (Kovacevic, Garcia and Gordillo-Tobar, 2023) and increasing concerns of the rise of AI technology, it is important to challenge current trends and study whether this increase in global military spending by governments is justified for the best interest of civilian welfare. This article therefore seeks to build upon the existing literature and examine the economic impact of defence spending on economic growth in the 21st century.

Literature review

Numerous studies have been conducted on the impact of defence spending on economic growth, with one of the first being carried out by Emilie Benoit in 1978. Benoit examined 44 less developed countries (LDCs) across 15 years (1950–65), and ultimately concluded that countries with the heaviest defence burden generally had the highest levels of growth (Benoit, 1978). Benoit used a basic multivariate linear regression model. He regressed civilian output growth (AG’) on the defence burden (AB) while controlling for investment rate (AI) and net foreign aid (AR).

From this model, Benoit found a strong positive correlation between the defence burden and economic growth. Benoit also highlights some possible reasons for this correlation, with him asserting that defence programs of most countries have a tangible effect on civilian economies. Some examples of these include the feeding and housing a number of citizens who would otherwise have been in poverty if it were not for conscription. Another theory posits that defence spending provides technical and medical training which may have a high civilian utility. Finally, Benoit also asserts that the construction of infrastructure such as roads/airports also serve civilian uses, and that the research and development that occurs from defence spending may have application to consumers/civilian industries.

Benoit’s results face criticisms, for example from Grobar and Porter. They researched the topic using a simple linear regression model between the ratio of military spending against the rate of GDP growth. Examining 44 LDCs between the years 1950–1965, they ultimately concluded that Benoit’s results of a strong positive correlation were not reproducible. Their possible reason for the differing results to Benoit is that Benoit’s results are not robust due to the data collection being from different countries using varying definitions of the variables incorporated. Their results suggest that whilst military spending can increase economic growth through some channels (e.g. technology spillover effects), they still posit that overall military spending slows down growth.The most significant net negative effect being the reduced national savings rates, which reduces the amount of capital available for investment.

Oğuz Taçyıldız and Asuman Çukur also explore the impacts of defence spending on economic growth in the 21st century, specifically focusing on Turkey. They tested the relationships between growth and two measures of defence spending, one being the measure from the Stockholm International Peace Research Institute (SIPRI) and the other being national budget defence expenditures (NBDE) from fiscal budget data.

Using a Granger causality test, they found a positive bidirectional relationship between defence SIPRI measures of defence expenditures and economic growth in Turkey. They conclude increases in defence expenditure defined by SIPRI lead to increases in economic growth, and that increases in economic growth also lead to increased defence spending. This differs from the result for NBDE figures, in which a positive unidirectional relationship between NBDE figures and defence spending was found. The results show a one way causality relationship between economic growth to defence expenditure as measured by NBDE, with economic growth causing increases in defence spending. The research found no statistically significant causal relationship between defence expenditure and economic growth.

Shoukat provides a further examination of the theoretical frameworks in which to analyse economic growth, and also provides an overview of the research which supports the different mechanisms. Within their work the author highlights common frameworks used to analyse the economic impacts of military expenditure are; demand, supply, and security (Shoukat, 2023). The author suggests that through the demand channel, military spending can impact economic growth through the Keynesian aggregate demand multiplier effect when there is spare capacity in the economy. Some recent empirical studies supporting this positive demand side impact include: Wijeweera and Webb, 2009 and Tiwari and Shahbaz, 2013. On the supply-side, these involve positive changes in the availability of factors of production in a country. For example, this could be technological spillovers from military research and development, with military innovations having civilian applications (e.g. nuclear energy, jet engines).

Examining the security channel, increased national security resulting from military spending can enhance the safety of individuals and property from threats, which encourages economic transitions and investment. One study which supports this thesis is by J Paul Dunne, R. P. Smith, and Willenbockel in 2005. They concluded that to some extent military spending can increase national security which can in turn facilitate economic activity and growth. This may be particularly relevant to developing countries, as major obstacles to development can include wars and lack of national security which prevent the development and integration of markets. Overall, they highlight that the influences of defence spending on the economy depends on the initial economic situation of the country investigated.

Methodology

Considering that Benoit’s paper is one of the first notable works within defence economics literature, added with its relatively direct approach, we would like to recreate his model with a few tweaks. Benoit’s original model is specified below:

Where it is regression of average growth of civilian product (AG’) on average defence burden (AB) and controlling for the average investment rate (AI) and average net foreign aid (AR). An error term ε is included to represent all other factors (i.e. shocks) that could influence a country’s civilian product. Benoit chose to average his data from 44 LDCs for all the variables (AG’, AB, AI, AR) over a period of 15 years — from 1950 to 1965. His results are as follows:

Based on Benoit’s regression, we can observe that a 1 unit increase in the average defence burden of low income countries, results in a 3.62 unit increase (mean) in average civilian product growth. We would like to observe whether these results can be replicated and are statistically significant enough to make any conclusions about the observed relationship.

For our research, we chose a different specification by including interaction variables and different independent variables. We also expanded our observations to all countries with available data and classified them into high, middle, and low income countries, based on theWorld Bank classifications. The least developed countries (LDCs) during the time of Benoit’s paper could also have since moved out of that category into either middle income or high income brackets, thus the regression results for countries in the low income category differ from Benoit’s as they consist of a different set of countries.

We believe that this distinction in the level of income as a proxy for a country’s level of development is important as we hypothesise that countries with higher levels of income would be more willing to spend a higher percentage of its GDP (defence burden) in defence of its geopolitical interests compared to a lower income country, and thus the effects of this spending on its economy may differ. However, it is also possible that lower income countries come with weaker institutions and may be more susceptible to civil wars and conflict thus positively driving its defence burden. These effects on the economy would also be interesting to understand with the aim of curating specific policies targeted at developing their economies.

For our regression, we utilise data from the Stockholm International Peace Research Institute (SIPR)I and World Bank databases, covering the years 2000 to 2022 (most recently available data). Note that we will be using average GDP growth (AG) figures from the World Bank instead of civilian product growth (AG’), as the latter’s data is not readily available. We also believe it would not make a huge difference as defence spending is typically a small fraction of GDP as a whole, thus when averaged, including it or not in our GDP growth data would not mean a large difference.

Further, gross capital formation as a percentage of GDP was used as a proxy for investment rate (AI), whereas net development aid as a percentage of GDP (ODA) was used to represent net foreign aid (AR). All the data was averaged over a 22 year period — emulating Benoit’s approach, as well as ruling out the need for more complex, dynamic regression models.

The total sample size for our dataset is 109 countries — these are the countries which have been consistently measuring our relevant variables without significant gaps in them. As a summary, we have 43 high income countries, 58 middle income countries and 8 low income countries. These classifications are based on the World Bank’s data.

With that, our methodology is described as a regression of average GDP growth (AG) on average defence burden of middle income countries (AB), average investment rate (AI), average net foreign aid (AR), average defence burden of high income countries relative to middle income countries (HI*AB) and average defence burden of low income countries relative to middle income countries (LI*AB).

In discerning the effects between countries with different income groups, we use middle income countries as the reference group (variable AB and coefficient β1) and incorporate interaction variables HI*AB and LI*AB which are the effects of a unit change in average defence burden for high income and low income countries respectively on the average growth of GDP. Dummy variables HI = 1 and LI= 0 represent the effects for high income countries and vice versa for low income countries. Thus, coefficients β4 and β5 are interpreted relative to any effects for middle income countries. Interaction variables were used to discern the effects between income levels without running the regression separately as well as controlling effects for all countries.

Therefore, a 1 unit change in average defence burden (AB) for middle income countries would result in a β1 unit change in average GDP growth (AG). On the other hand, a 1 unit change in average defence burden (AB) for high income countries would lead to a β1 + β4 unit change in average GDP growth (AG). The same would ring true for low income countries where a 1 unit change in AB leads to a β1 + β5 unit change in AG.

Analysis

These are the results from running the regression in Stata:

Focusing on the effect of defence spending for middle income countries (reference group, ab), we observe that a 1 unit increase in the average defence burden is associated with a 0.29 unit increase in average economic growth, and that this result is statistically significant at the 5% significance level (t = 2.13 > 1.96).

For the case of high income countries (hiab), we can observe that a 1 unit increase in the average defence burden leads to a decrease of -0.039 units of average GDP growth, and that this result is statistically significant (t = 2.57 > 1.96). This result is calculated by deducting 0.3298 from the coefficient for middle income countries, 0.2908. Therefore, we conclude that there is a negative relationship between average defence burden and average GDP growth for high income countries.

Lastly, in the case of low income countries (liab), a 1 unit increase in average defence burden leads to a 0.5345 unit increase in average GDP growth. It should be noted that this result mirrors that of Benoit’s with a substantial positive correlation coefficient. However, when tested at the 5% significance level, this result cannot be deemed statistically significant (t = 0.61 < 1.96). We therefore fail to reject the null that there is no relationship between average defence burden and average GDP growth for high income countries.

Overall, when running our regression using 21st century data, closely following Benoit’s specification, we can conclude that his results for low income countries, with their being a positive correlation between defence spending and civilian product cannot be replicated as the result is not statistically significant. However this may be due to our small sample size of 8 low income countries. The result is, however, statistically significant for middle and high income countries with high income countries exhibiting a negative correlation coefficient.

We believe that this difference in the sign of the correlation coefficients for the 3 income categories is due to a number of reasons. For high income countries, the negative correlation could be due to crowding out effects which refers to when high levels of defence leads to a reduction in private sector spending (Anderton et al, 2009). Military projects will often require a significant amount of resources (raw materials, skilled labour, capital goods etc). Thus, increased defence spending by the government will increase demand for these resources and drive up prices, meaning that they are less affordable for the private sector. This increased cost of factor inputs could thus crowd out private sector investment, negatively impacting consumer production. Further, increased military spending can also displace important government investment from more productive industries (i.e., healthcare, education, R&D) impeding a nation’s long-term economic growth (Dunne, Pieroni, et al, 2017).

In the case of middle income and low income countries, the positive effects could be due to crowding in and growth spin-offs (Anderton et al, 2009). The crowding in effect is opposite to crowding out with government defence spending causing an increase in private sector investment. Increased defence spending from the government could raise the incomes of workers in military related sectors. This increase in incomes for certain workers could in turn stimulate demand for consumer goods, spurring economic growth.

Growth spin-offs refer to secondary economic effects resulting from military spending. These can often contribute to economic growth. Technology initially developed for military use may later have consumer/industrial applications (e.g. military communication, aerospace technologies). Thus, if technology initially created for military purposes can later be applied to consumer industries, this can lead to economic growth. The increased sense of safety within low income countries resulting from increased defence spending could also contribute to a better sense of trust in economic institutions, spurring economic growth.

Limitations of the model

We proceed to discuss the limitations of recreating Benoit’s model. For one, the dataset may not be accurate as low-income countries such as Afghanistan, Sudan, and Syria report missing or inaccurate values due to a lack of sophisticated data collecting capabilities compared to those in high-income countries. This is especially unfortunate as it is countries such as these that are in war-torn areas thus, it would be interesting to understand how these conflicts affect the economy and its people.

Further, using averaged data may not reflect the potential fluctuations throughout the 21 year period especially during periods of shocks (i.e., 9/11 attacks, 2008 Financial Crisis, Covid-19 pandemic). These particular events could have some impact on the defence spending of countries — for instance the 9/11 attacks had a profound effect on global defence spending particularly in the US where military spending spiked 50% in the decade following the attacks. These events acting as exogenous shocks (in the error term epsilon) affect not only our observed variable but also our dependent variable in the regression. Regardless, in our regression and Benoit’s the averaging of our data was done to avoid the complications involved in using panel data.

Another important limitation is the potential for confounders — other variables outside our regression that are correlated to our independent variables (i.e., political and geopolitical factors, inflation, government debt levels) and thus could impact our dependent variable indirectly. Thus, one of the means to correct this is to include further controls in order to partial out the effects these confounders have on our regression.

Overall, we realise that there are significant limitations in our recreation of Benoit’s method as discussed above. However, we do find it worthwhile to attempt a recreation of his work, a titular piece of writing on defence economics, as this is a means of understanding what a rise in global defence spending amidst the conflicts occurring in our world today.

Conclusion

Overall, the results of our regression analysis and literature review are nuanced. The relationship between defence spending and economic growth is difficult to discern and dependent on the level of income of a country and its current economic condition, with middle and low income countries possibly benefiting from an increase in defence spending whereas high income countries experiencing otherwise.

There are a myriad of factors impacting the result which include global shocks such as the 9/11 attacks, 2008 Financial Crisis, Covid-19 pandemic and European energy crisis, that act as confounders in our regression, making identification unfeasible . However, we still believe that such investigation is essential as a means of guiding governments around the world to consider not only the social costs of war and increased military spending, but also the economic ones.

This article was written by Meisha Lukman and Ivan Beales, Economic Research Analysts in the Economic Research Division. This article was edited by Cai Hui Lien, Head of the Economic Research Division.

Bibliography

Benoit, E. (1978). Growth and Defense in Developing Countries. Economic Development and Cultural Change, 26(2), pp.271–280. doi:https://doi.org/10.1086/451015.

Çukur, A. and Taçyıldız, O. (2022). The Impact of Defense Expenditures on Macroeconomic Indicators in Turkey: 2000–2020 Period. Journal of Security Studies, 24(1), pp.50–74.

Deger, S. and Sen, S. (1995). Military expenditure and developing countries. In: K. Hartley and T. Sandler , eds., Handbook of Defense Economics. Elsevier, pp.275–307.

Dunne, J.P., Smith, R.P. and Willenbockel, D. (2005). Models of Military Expenditure and Growth: a Critical Review. Defence and Peace Economics, 16(6), pp.449–461. doi:https://doi.org/10.1080/10242690500167791.

Goddard, T. (n.d.). Guns Before Butter. [online] politicaldictionary.com. Available at: https://politicaldictionary.com/words/guns-before-butter/ [Accessed 13 Jan. 2024].

Grobar, L.M. and Porter, R.C. (1989). Benoit Revisited: Defense Spending and Economic Growth in LDCs. The Journal of Conflict Resolution, 33(2), pp.318–345.

H. Anderton, C. and R. Carter, J. (2009). Principles of Conflict Economics: A Primer for Social Scientists. Public Choice, 145(1–2), pp.15–27. doi:https://doi.org/10.1007/s11127-010-9667-9.

Kovacevic, R., Garcia, J.N.B. and Gordillo-Tobar, A. (2023). With mental health conditions on the rise, countries must prioritize investments. [online] blogs.worldbank.org. Available at: https://blogs.worldbank.org/health/mental-health-conditions-rise-countries-must-prioritize-investments [Accessed 21 Jan. 2024].

Marrow, A. (2023). Russia’s budget deficit seen at around 1% of GDP this year -finance minister. Reuters. [online] 22 Nov. Available at: https://www.reuters.com/markets/europe/russias-budget-deficit-seen-around-1-gdp-this-year-finance-minister-2023-11-22/ [Accessed 17 Jan. 2024].

Masterson, V. and North, M. (2023). What is ‘global debt’ — and how high is it now? [online] World Economic Forum. Available at: https://www.weforum.org/agenda/2023/12/what-is-global-debt-why-high/.

Shahbaz, M., Afza, T. and Shabbir, M.S. (2013). Does defence spending impede economic growth? Cointegration and causality analysis for Pakistan. Defence and Peace Economics, 24(2), pp.105–120. doi:https://doi.org/10.1080/10242694.2012.723159.

SIPRI (2023). World military expenditure reaches new record high as European spending surges | SIPRI. [online] www.sipri.org. Available at: https://www.sipri.org/media/press-release/2023/world-military-expenditure-reaches-new-record-high-european-spending-surges.

Shoukat, H. 2023. Essays on economic growth, military expenditure, armed conflict and corruption PhD thesis, University of reading. doi:https://doi.org/10.48683/1926.00113453

Tiwari, A.K. and Shahbaz, M. (2013). Does Defence Spending Stimulate Economic Growth in India? A Revisit. Defence and Peace Economics, 24(4), pp.371–395. doi:https://doi.org/10.1080/10242694.2012.710814.

United Nations Development Program (2022). 6 in 7 people worldwide plagued by feelings of insecurity | United Nations Development Programme. [online] https://www.undp.org/. Available at: https://www.undp.org/european-union/press-releases/6-7-people-worldwide-plagued-feelings-insecurity [Accessed 21 Jan. 2024].

Wijeweera, A. and Webb, M.J. (2009). Military Spending and Economic Growth in Sri Lanka: A Time Series Analysis. Defence and Peace Economics, 20(6), pp.499–508. doi:https://doi.org/10.1080/10242690902868301.

Zurcher , A. (2023). US House passes controversial defence spending bill. BBC News. [online] 14 Jul. Available at: https://www.bbc.co.uk/news/world-us-canada-66206542 [Accessed 21 Jan. 2024].