Military spending crowding out health and education spending: which views are valid in Egypt?

The paper investigates whether government spending components—military, health, and education—drive or respond to economic growth in Egypt (1980–2021), contrasting Wagner’s Law (spending is endogenous to growth) with the Keynesian view (spending is exogenous and stimulates growth). Egypt’s military expenditure has fluctuated due to conflicts, debt relief, and political upheavals, notably peaking around 2011 during the uprising and subsequent security operations. Given Egypt’s history of political conflict, the study posits that prioritizing military spending may have detracted from human development. Education’s share in government spending peaked at 3.3% in 2010, dropped markedly during 2011–2014, and then recovered by 2017; health spending has been relatively low (averaging 3.7% of GDP), with declines around 2010–2016 before increasing. Prior Egyptian studies often used aggregate spending and did not examine bidirectional causality or the impact on specific components like health and education. This study contributes by testing Wagner vs. Keynes in a disaggregated framework and by examining whether military spending crowds out health and education, including the direction of causality among these components and GDP.

Theoretical background: Wagner (1958) posits a unidirectional causality from economic growth to government spending, implying spending expands with development; this has been formalized and tested in subsequent work (Goffman and Mahar, 1971; Peacock and Wiseman, 1961; Akitoby et al., 2006; Kibara Manyeki and Kotosz, 2017; Uppal and Glazer, 2015). Keynes (1937, 1963) views government spending as an exogenous tool to stimulate demand and output; Barro (1990) adds that government expenditure can influence growth via production functions in an endogenous growth setting. Military Keynesianism argues defense outlays stimulate demand, investment, and employment (Atesoglu, 2002; Nordhaus, 2002), though critics (Melman, 1978; Dunne, 2012; Hossein-zadeh, 2009; Gold, 2005; Kinsella, 1990; Payne and Ross, 1992) highlight negative or insignificant growth effects and potential crowding out of human and physical capital. A key critique of prior empirical work is reliance on aggregate spending, which masks allocation effects across sectors like health and education (Dritsakis and Adamopoulos, 2004; Gupta, 1969; Hondroyiannis and Papapetrou, 1995). Empirical review: Studies using ARDL and cointegration often support Wagner’s Law in various countries (Islam, 2001; Burney, 2002; Narayan et al., 2008; Samudram et al., 2009; Atilgan et al., 2017; Kesavarajah, 2012; Magazzino, 2012; Manamperi, 2016; Karagianni and Pempetzoglou, 2009; Shelton, 2007), while others test both Wagner and Keynes jointly (Paparas et al., 2019; Ghazy et al., 2021). Military Keynesian frameworks have been explored via Granger causality/VAR (Dunne and Vougas, 1999), IS-LM structural models (Atesoglu, 2002; Pieroni et al., 2008; Smith and Tuttle, 2008), and augmented production functions (Dunne and Nikolaidou, 2005), with mixed causality results. For Egypt and conflict regions, Abu-Bader and Abu-Qarn (2003) examined civilian vs. military spending without disaggregating civilian items; Ali (2011) found inconclusive military spending–growth links; Eldemerdash and Ahmed (2019) tested aggregate Wagner vs. Keynes and did not find causality from spending to GDP. Recent literature on crowding out shows military outlays can reduce education quality (Jesmy et al., 2015) and may hinder growth in non-OECD countries (Azam, 2020), though some panel evidence finds positive growth effects (Raifu and Aminu, 2023). On health, military spending can raise healthcare demand but crowd out health budgets (Ikegami and Wang, 2023), with mixed findings elsewhere (Biscione and Caruso, 2021; Coutts et al., 2019). Decomposition studies suggest personnel spending is particularly growth-damaging (Becker and Dunne, 2023). The literature lacks consensus on Wagner vs. Keynes for military spending and rarely examines how military spending affects allocations to health and education, especially in Egypt—motivating this study’s disaggregated, causal analysis.

Scope and data: Annual data for Egypt, 1980–2021, from World Bank Indicators, UNICEF, SIPRI Military Expenditure Database, UNESCO, and Egypt’s Ministry of Finance. Missing observations for health and education variables were interpolated. Variables were transformed into natural logs. The study adopts the Pryor (1968) Wagner specification and follows Al Qudah et al. (2020) for empirical strategy. Empirical steps: (1) Descriptive statistics and correlation analysis; multicollinearity checked via variance inflation factor (VIF), with VIF<10 threshold (Alin, 2010). Reported VIFs: LMIL 4.2141; LGOVHEL 8.5524; LGOVEDU 6.4155, indicating no serious multicollinearity. (2) Stationarity tested using Augmented Dickey–Fuller (ADF) and Phillips–Perron (PP) tests. At levels, LGDP, LMIL, LGOVHEL, and LGOVEDU are non-stationary; first differences are stationary (ADF/PP reject unit roots at 10% or 5% levels depending on series). (3) Structural breaks assessed with Chow tests, identifying breaks at 1990 and 2011 across variables (Fisher, LLR, and Wald statistics all significant at 1%). (4) Optimal lag length determined via information criteria (AIC, FPE, etc.), selecting 1 lag. (5) Cointegration assessed using Pesaran–Shin–Smith bounds approach: no cointegration at levels (statistic −7.3279, p=0.9827); cointegration confirmed in first differences (statistic −28.41, p=0.0000). (6) Short-run causality examined via Granger causality tests with two lags. (7) Dynamic relationships estimated with ARDL integrated with an Error Correction Model (ECM) to obtain short- and long-run effects on LGDP from LMIL, LGOVHEL, and LGOVEDU. Model fit reported with R-squared; adjustment speed gauged by lagged error-correction term.

• Granger causality (Table 6): Unidirectional causality from spending components to GDP, supporting the Keynesian direction and rejecting Wagner’s Law for Egypt. Specific results: LMIL → LGDP (F=8.9655, p=0.0021); LGOVHEL → LGDP (F=7.1045, p=0.0052); LGOVEDU → LGDP (F=12.5201, p=0.0001). No causality from LGDP to these components. - Inter-component causality: LMIL → LGOVHEL (F=9.4104, p=0.0002) and LMIL → LGOVEDU (F=8.0252, p=0.0014), indicating military spending influences health and education allocations. No reverse causality from health or education to military spending. Education → health also shows significance (LGOVEDU → LGOVHEL: F=10.1102, p=0.0014), while the reverse is not significant. - ARDL-ECM long-run coefficients on LGDP (Table 7): LMIL −6.982 (p=0.000), LGOVHEL 2.433 (p=0.001), LGOVEDU 5.334 (p=0.012). Interpretation: 1% increase in military spending associates with a 6.98% decrease in GDP in the long run; health and education spending raise GDP by 2.43% and 5.33%, respectively. - ARDL-ECM short-run coefficients (Table 7): DLMIL −0.093 (p=0.005), DLGOVHEL −0.062 (p=0.042), DLGOVEDU −0.045 (p=0.001). Interpretation: Increases in military, health, and education spending reduce GDP in the short run, consistent with adjustment lags in human capital formation and potential short-term crowding-out effects. - Error-correction term: Lagged residual (−1) = −0.0365 (p=0.089), implying a slow adjustment to long-run equilibrium at about 0.36% per period. - Cointegration and breaks: No cointegration at levels; cointegration in first differences. Structural breaks identified at 1990 and 2011, consistent with ERSAP reforms and the 2011 revolution. - Correlations: Military spending negatively correlated with GDP (−0.8255) and with health spending (−0.4520), positively with education spending (0.9125), though allocation opacity limits interpretation. - Model fit: R-squared ≈ 59.63% for the ARDL-ECM model.

The study’s primary question—whether government spending components drive growth (Keynes) or respond to it (Wagner)—is addressed by Granger causality results showing a one-way impact from military, health, and education spending to GDP, rejecting Wagner’s Law for Egypt. However, despite this Keynesian direction, the estimated effects reveal that military spending reduces GDP in both short and long horizons, aligning with arguments that defense outlays can crowd out productive investment and human capital and may impose future burdens. In contrast, health and education spending have negative short-run effects—consistent with adjustment lags and implementation frictions—but significant positive long-run effects on GDP, supporting endogenous growth mechanisms through human capital, innovation, and productivity. The causality from military to health and education budgets suggests that defense priorities shape social sector allocations, with evidence of a negative association with health and a positive association with education potentially reflecting military-affiliated education systems. Overall, the findings nuance the Keynesian perspective: while spending is exogenous to growth, not all components are growth-enhancing; composition matters critically, and military spending appears growth-dampening in Egypt’s context.

The analysis for Egypt (1980–2021) rejects Wagner’s Law and supports a Keynesian direction of causality from spending components to GDP. Yet, compositionally, military spending is detrimental to GDP in both the short and long term, while health and education spending hinder growth in the short term but significantly bolster it in the long term. Policy implications include reducing military outlays and prioritizing health and education to foster sustainable growth via human capital, innovation, and productivity. The results also indicate that military spending influences social sector allocations, warranting greater transparency and granular data on defense budgets, especially education-related components. Future research should incorporate infrastructure spending, explore cross-country comparisons (e.g., with South Africa), and leverage more detailed budget decompositions to fully capture allocation dynamics and their growth impacts.

• Data gaps for health and education were addressed via interpolation, potentially introducing measurement error. - Lack of transparent, granular data on military budget allocations (e.g., personnel vs. capital, military education components) constrains identification of specific channels and the interpretation of the positive association with education spending. - Single-country time-series design (Egypt, 1980–2021) limits generalizability; structural breaks (1990 ERSAP, 2011 revolution) may affect parameter stability despite Chow tests. - Slow error-correction speed suggests possible model misspecification or omitted variables (e.g., infrastructure, investment, external shocks), which future work could address.