Several contemporary empirical studies have been conducted with the same perspective to explain this negative relationship in various contexts. The main ones are presented below. [6] examined the impact of military spending on economic growth in non-OECD (Organization for Economic Co-operation and Development) countries. The results clearly show a negative effect of this spending on growth, as it does not contribute favorably to its determination. Similarly, [27], [40], and [33] observed a negative and statistically significant impact of military spending on economic growth in many countries. [15] also highlighted the adverse consequences of such spending in developing countries.

Conversely, some results prove positive. [64] also provide empirical insights into the positive relationship between military spending and economic growth in the Middle Eastern countries and as a whole. Through cointegration tests, the authors concluded that there is a positive long-term relationship between military spending and gross domestic product. Building on this, [53] assesses the case of Türkiye over the period 1956-1994 by simultaneously estimating several interdependent economic relationships (eg, between growth, military spending, savings, and the trade balance). The results explain that military spending has a positive effect on overall economic growth but does not have a significant impact on certain macroeconomic aggregates such as savings and the trade balance. Finally, [12] analyzed the impact of military spending on the Economic growth in developing countries using a cross-sectional approach from 1950 to 1965. The results showed a positive correlation between military spending and GDP growth. This refers to a multiplier effect of public spending, including military expenditures.

The economic literature identifies several determinants of military spending, with economic growth playing a central role. Numerous empirical studies indicate that increased national income and fiscal capacity tend to encourage higher defense budgets ([55]; [26]; [11]). Beyond economic factors, other major determinants have been highlighted in the literature, namely security threats, geopolitical factors, and institutional quality. For example, [41] and [49] explain the role of regional tensions and military alliances, while [45] emphasizes the influence of the international security environment. [2], for their part, find that institutional characteristics and governance play a decisive role in the allocation of military spending.

In the African context, [58], in a study covering 12 Southern African countries between 1997 and 2004, showed that GDP per capita is a major explanatory factor for military spending. Similarly, [9], using South African data, highlight a positive correlation between income level and military allocations. This trend is also observed in China, where [57] established that Gross National Product growth was a significant driver of military spending.

The study covers eight WAEMU countries (Benin, Burkina Faso, Côte d’Ivoire, Guinea-Bissau, Mali, Niger, Senegal, and Togo) over the period 2000-2023, representing 192 annual observations. Data on variables such as economic growth, population, financial development, remittances, trade openness, rule of law, political stability, democracy, regulatory quality, and foreign direct investment (FDI) are from the World Bank. Military expenditure data come from the SIPRI (Stockholm International Peace Research Institute) database. Research Institute.

Growth economic (GDP per capita) is measured by the GDP per capita growth rate in this study. This choice East based on the study recent work by [50]. The variable “Expenditure” military spending public allocated to the security sector, including the acquisition equipment, recruitment, staff remuneration and investments in research and development, expressed as a percentage of GDP. This choice of variable is similar to that of [50] and [52]. Financial development, measured in this study by the financial credit granted by banks to the sector private serves as a transition variable. According to the results of [22]; a sign positive East expected in this article. The opening trade, measured by the ratio of trade value trade on GDP, that is, the sum of exports and imports and GDP. This choice East Finally motivated by the study of [52]. We expect a positive relationship between growth rates economic and open. The population is approximated by the population growth rate (POP) as in [1]. This variable is chosen in accordance with [46]. Political stability and the absence of violence correspond to the assessment of the risk of destabilization. or the overthrow of the government, that this either by means institutional or by the use of force, this choice East motivated in the literature by the study of [46]. The democracy index (demo) is retained in the study to capture the quality of institutions, in accordance with [30], a sign positive East expected in this work. Regulatory quality (RQ) encompasses policies that could disrupt the proper functioning of the market, including measures such as price controls. [4] show that this variable has an effect positive on spending military, a sign positive East expected in this work. The rule of law encompasses all laws and regulations established in a country in order to ensure the protection of the interest general and to structure the daily lives of citizens. Following [46], we choose this variable. The IDE East a shape transnational investment in which a resident of one country holds controlling power or a significant influence on the management of a business located in another country. [62] uses this variable to analyze its effect on growth in ECOWAS. The authors found an effect positive impact of FDI on growth Economically, we expect a sign positive in this article. Net remittances from migrants (RM) are evaluated in adding up workers’ remittances, employees’ remuneration as well as capital transfers made by migrants. [22] found an effect positive on growth economy. Table 1 presents the description of the variables and the data sources.

Table 1. Variable summaries and sources.

Source: Literature.

Table 2. Descriptive statistics results.

Source: Author, WDI and SIPRI data.

Table 2 presents the descriptive statistics. The average military expenditure is 11.59%, with a high standard deviation (6.91) reflecting significant heterogeneity between countries. The minimum (1.66%) and maximum (35.54%) suggest substantial differences, likely linked to specific security contexts. Growth shows a low average of 1.45% with marked variability (standard deviation of 2.77).

VIF Test

Since we are dealing with several variables, we will also perform a multicollinearity test of the variance inflation factor (VIF). The VIF assesses whether the factors are correlated with each other (multicollinearity), which could influence other factors and reduce the model’s reliability. If the VIF is greater than 10, we have high multicollinearity: the variation will appear larger, and the factor will appear more influential than it actually is. If the VIF is close to 1, then the model is much more robust because the factors are not influenced by correlation with other factors. Table 3 and Table 4 show an absence of multicollinearity in our data.

Table 3. Multicollinearity test based on VIF calculation model 1.

Source: Our calculations based on SIPRI and WDI data.

Table 4. Multicollinearity test based on the calculation of the VIF model 2.

Source: Our calculations based on SIPRI and WDI data.

3.2.1. Justification of the Methodology

We use Triple least squares (3SLS) estimator. This choice is based on several arguments. Firstly, it allows for better handling of endogeneity problems and mitigates heteroscedasticity. Secondly, the 3SLS estimator is thus more robust for estimating systems of simultaneous equations. Compared to two-stage least squares (2SLS) and the GMM of Arellano and Bond, 3SLS offers increased efficiency by reducing the simultaneity bias. Furthermore, [24] emphasizes that among the various instrumental variable-based methods, the 3SLS estimator is asymptotically efficient. This approach relies on a first step where each equation is estimated using two-stage least squares (or an instrumental variable method), followed by a second phase where the resulting residuals allow us to evaluate the correlation of errors between the different equations.

3.2.2. Model Specifications

Following the work of [17], we treat military spending and economic growth as jointly endogenous.

First equation: GDP per capita

According to the literature, our equation representing economic growth is specified as follows:

Or i represents the country, t the time dimension. DPM represents military spending, Xis a vector of control variables intended to influence growth. γ i represents the individual specific effect which can be fixed or random, μ t the time-specific effect and ε i , t the error term of the model.

Second equation: military spending

Empirical studies on the determinants of military spending have employed a variety of econometric techniques. Furthermore, these studies have also highlighted the potential influence of political, geographic, and socioeconomic factors on the structure and evolution of military spending. Finally, we specify a second equation in which military spending is explained separately by development.

Based on the empirical frameworks proposed by [2] and [60], the following empirical model is specified.

The empirical model

Finally, the simultaneous equations model that we retain consists of the two equations (1 and 2):

3.2.3. Identification: Order Condition

The order condition is being studied equation by equation. For each equation, we have:

GDP per capita equation

In this equation:

Endogenous included: GDP, DPM→ g ′ = 2

Exogenous included: POP, FDI, TRADE, RM, DF→ k ′ = 5

Let’s apply the ordercondition

Conclusion: 1 < 4, therefore the equation is over-identified.

Equation 2:

Endogenous included: GDP, DPM→ g ′ = 2

Exogenous included: POP, STABILITY, TRADE, RL, QR, DEMO, k ′ = 6

Conclusion: 1 < 3, therefore the equation is over-identified.

A. Dependency Test

The dependence test allows us to choose the unit root tests best suited to our study model. Since our study period is longer than the number of individuals, we will therefore use the Breusch-Pagan dependence test ([14]) to test the hypothesis of non-dependence between individuals.

The principle of the test is as follows:

H0: dependency;

H1: independence.

Table 5. LM independence test, Breuch-Pagan test.

Source: Author. Note: ***, ** and * indicate statistical significance at 1%, 5% and 10% level respectively.

The results of this test show that the p-value associated with the test statistic is less than 1%, therefore we reject the null hypothesis of independence. Consequently, the individuals in the study are interdependent. Given the presence of dependence between the individuals, it is therefore obvious to apply stationarity tests. To confirm our results, we also performed the Pagan LM test, and the results do indeed show the dependence between the individuals.

B. Stationarity Tests

The results of the dependency test (Table 5) indicate the presence of cross-dependency between our data; in this context, we use second-generation unit root tests, including CIPS and CADF, proposed by [47].

CIPS is defined by:

CADF expresses itself as follows:

Here, Y ¯ i t − 1 represents the lagged mean value, while Δ Y ¯ i t − 1 corresponds to the mean of the differentiated values for each series.

Table 6. Unit root tests.

Source: Author. Note: ***, ** and * indicate statistical significance at 1%, 5% and 10% level respectively.

Table 7. KAO test model 1.

Source: Author.

Cointegration test

The cointegration test will allow us to test for the existence of a long-term relationship between the variables in the model (Table 6). The ideal test would be the Pedroni test but against Given the large number of variables (more than 7), we cannot so do not use This test. Therefore, we will use the Kao test (Table 7).

C. Model estimation results

Table 8. Model estimation results.

Source: Author. Note: *, ** & *** denotes the significance 10%, 5% and 1% level respectively.

Statistically, several variables in the model explaining GDP per capita are significant. Military spending has a significant negative effect at the 1% level (p =0.003) (Table 8).

D. Causality analysis in the sense of[18]

The objective of this section is to present the results concerning the causal relationship between military spending and economic growth in the WAEMU. The results presented in Table 3 provide strong evidence of a bidirectional causal relationship (feedback relationship) between military spending (MSW) and economic growth. Indeed, the null hypothesis of no causality is rejected in all models, with panel test statistics being significant at the 1% level, which confirms the existence of a dynamic interaction between the two variables. These observations are consistent with the findings of [36], [34], [35]; [37]; [59]; and [51] in 12 African countries. Thus, each of these two variables influences the other: any change (up or down) in defense budgets has a proportional impact on the level of economic activity. For the countries concerned, these results indicate that defense and growth policies are designed to reinforce each other. They also appear to be closely linked, to the point that a change in one necessarily affects the other (Figure 1, Table 9).

Table 9. Causality results in the sense of [18].

Source: Author.

Source: Author.

Figure 1. Graphical representation of the causal relationship.

The significant negative impact of military spending on economic growth, measured here by GDP per capita, confirms a thesis widely supported by the literature on developing savings. Funds allocated to military spending represent resources not invested in productive sectors such as education, health, or economic infrastructure ([8]). This finding aligns with the neoclassical view that military spending is unproductive or, at best, makes little contribution to long-term economic growth. These results on military spending confirm those previously obtained by [42], [21], [54], and [20]. Furthermore, the negative impact of population on economic growth can be attributed to increased pressure on natural resources, infrastructure, and public services, consistent with the predictions of [43]. In WAEMU countries, rapid population growth can make it difficult to improve living standards, especially if investments in education, health, and infrastructure do not keep pace. This finding contradicts that of [62], who found a positive effect of population growth on economic growth in ECOWAS. However, this result aligns with that of [50] in Africa and [46]. These authors found a negative effect of population growth on economic growth. The positive effect of financial development on economic growth is positive, which is consistent with the literature showing that well-developed financial systems facilitate access to credit, reduce transaction costs, and improve resource allocation. The study’s findings confirm those of previous work by [22], [39], and [10].

Conversely, the results indicate that economic growth has a positive and significant impact on military spending. Specifically, a 1% increase in growth, all other things being equal, translates into a 1.66% increase in military spending as a percentage of GDP in WAEMU countries. According to [19], higher per capita income stimulates aggregate demand, thereby encouraging the hiring of new workers and the expansion of investment. This finding aligns with the conclusions of [16], who observed a positive and significant effect on military spending using GDP per capita. The positive and significant coefficient for regulatory quality suggests a direct correlation between a well-regulated institutional environment and increased military spending. This relationship can be explained by the fact that strong regulatory institutions facilitate better management of public finances, including those related to defense. This aligns with the argument put forward by [4] that improved institutional governance can paradoxically strengthen the state’s coercive capabilities, even within a democratic framework. The positive effect of trade openness on military spending can be interpreted through the lens of economic security. Greater openness exposes countries to external risks (piracy, cross-border insecurity, strategic dependence), thus justifying increased spending to protect economic interests. This finding is consistent with [32] conclusions that integration into the global economy increases security needs, particularly for securing trade corridors and strategic infrastructure.