## wpiea2019196-print-pdf - References

## Source details

**Canonical URL:** [wpiea2019196-print-pdf - References](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019196-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019196-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019196-print-pdf.pdf.json)

---

### I. Introduction and key trends
- Worldwide military spending (unweighted country averages) declined from 3.6 percent of GDP during the Cold War period (1970-1990) to 1.9 percent of GDP during 2010-2018.
- The share of the budget allocated to defense fell by three percentage points between the Cold War period and 2010-2018.
- Disaggregation:
  - Developing countries experienced a sharper fall in military spending as a share of total spending than advanced economies.
  - Total global military spending annual averages:
    - US$410,927 million during 1970-1990.
    - US$1,703,810 million during 2010-2018.
- Regional patterns:
  - Middle East average military spending: 4.5 percent of GDP (1992-2018 period).
  - Western Hemisphere average: around 1.4 percent of GDP.
  - Sub-Saharan Africa averaged 2.1 percent of GDP for the entire period.
  - Nearly half of Sub-Saharan African countries are classified as fragile states.
- Distribution of countries by military spending (1970-2018):
  - 57 countries with military spending under 2 percent of GDP.
  - 68 countries with military spending between 2-5 percent of GDP.
  - 13 countries with military spending in excess of 5 percent of GDP.
  - 100 countries allocate less than 10 percent of the budget to the military.
  - 31 countries allocate between 10 and 20 percent of the budget to the military.
  - 7 countries allocate more than 20 percent of the budget to the military.
- Several advanced economies are among the top 15 military spenders and account for over 80 percent of worldwide military spending.

### II. Summary statistics (Table 1)
- Military expenditure as percent of GDP — means and standard deviations by period and group:
  - All:
    - Mean: 1970-1990 = 3.61; 1992-2000 = 2.79; 2001-2009 = 2.10; 2010-2018 = 1.90
    - St.Dev.: (3.79) (3.09) (1.80) (1.60)
  - Advanced:
    - Mean: 1970-1990 = 3.60; 1992-2000 = 2.23; 2001-2009 = 1.85; 2010-2018 = 1.59
    - St.Dev.: (3.80) (1.72) (1.29) (1.01)
  - Developing:
    - Mean: 1970-1990 = 3.62; 1992-2000 = 2.98; 2001-2009 = 2.18; 2010-2018 = 1.99
    - St.Dev.: (3.72) (3.40) (1.93) (1.73)
  - Sub-Saharan Africa:
    - Mean: 1970-1990 = 2.61; 1992-2000 = 2.78; 2001-2009 = 1.87; 2010-2018 = 1.67
    - St.Dev.: (2.27) (4.44) (2.18) (1.27)
- Military expenditure as percent of total expenditure — means by period (All):
  - Mean: 1970-1990 = 9.58; 1992-2000 = 9.12; 2001-2009 = 7.46; 2010-2018 = 6.42
  - Advanced means: 6.98, 5.31, 5.07, 4.33 (1970-1990; 1992-2000; 2001-2009; 2010-2018)
  - Developing means: 12.58, 10.54, 8.20, 7.07 (same periods)
- Sources: SIPRI, IMF World Economic Outlook.

### III. Explanatory factors cited in the literature
- Factors proposed to explain the decline in military spending:
  - End of the Cold War and improved global security environment.
  - Fiscal consolidation pressures in advanced economies, with average debt-to-GDP ratio exceeding 100 percent in the post-financial crisis period (IMF 2018).
  - Developing countries shifting budget shares toward social sectors and infrastructure since the early 2000s, driven by growth and SDG needs.
  - Aging populations in advanced economies increasing pension and health spending pressures (Clements et al. 2017).
- Prior literature and methods referenced:
  - AKA (Gleditsch et al. (1996), Gupta, Schiff, and Clements (1998), Rockoff (1998), Avranitidis, Kollias and Anestasopoulos (2014)) studied sigma and beta convergence for 128 countries during 1989-2008.
  - Lau, Demir, and Bilgin (2015) used nonlinear unit root tests on 37 countries (1988-2012) and analyzed pairwise convergence thresholds.

### IV. Convergence testing: data and methods
- Data:
  - SIPRI annual military expenditure data for 140 economies (34 advanced economies, 106 developing economies) covering 1970-2017.
  - SIPRI includes current and capital spending on armed and paramilitary forces, government agencies engaged in defense projects, and military space activities; expenditures on personnel (including pensions), operations and maintenance, procurement and construction, research and development, civil defense, and demobilization.
- Tests applied:
  - Sigma convergence (cross-sectional variance over time) shows convergence for advanced economies, developing countries, and the world as a whole (Figure 4).
  - Cross-sectional beta convergence baseline specification (dependent variable: average annual change in military expenditure as percent of GDP over T; independent variable: initial y).
    - Table 2 results (Cross-Sectional Beta Convergence Test):
      - Coefficient on y_0:
        - All: -0.0237*** [0.0018]
        - Advanced: -0.0250*** [0.0012]
        - DE: -0.0236*** [0.0023]
      - Cons:
        - All: 0.0364*** [0.0082]
        - Advanced: 0.0277*** [0.0051]
        - DE: 0.0400*** [0.0106]
      - Obs.: 140 (All); 34 (Advanced); 106 (DE)
      - R-Sq: 0.56 (All); 0.93 (Advanced); 0.51 (DE)
    - Interpretation: For the whole sample the point estimate of the beta coefficient is -0.02; a one-percentage higher initial value is associated with a 0.2 percentage faster decrease in military expenditure as a percentage of GDP.
  - Phillips and Sul (2007) club convergence methodology:
    - Decomposes y_it = phi_i * mu_t + eps_it and constructs relative transition coefficients h_it and cross-sectional variance V_t.
    - Tests convergence by estimating log(V_1) - log(V_t) - 2 log L(t) = a + b log t + eps_t for t = rT,...,T and testing whether b >= 0 (equivalently alpha >= 0).
    - Practical modeling uses delta_it = delta_i + sigma_i * theta_it * L(t) * t^(-alpha).

### V. Convergence group classification and patterns (Table 3, Figures 5–6)
- Convergence group test results (Table 3):
  - Group 1:
    - Countries: 13
    - Time periods: 49
    - Coefficient: 0.55**
    - S.E.: 0.35
    - t-stat.: 1.58
  - Group 2:
    - Countries: 98
    - Time periods: 49
    - Coefficient: -0.02**
    - S.E.: 0.01
    - t-stat.: -1.60
  - Group 3:
    - Countries: 27
    - Time periods: 49
    - Coefficient: -0.19**
    - S.E.: 0.23
    - t-stat.: -0.82
  - Note: The null hypothesis is that all countries within each group converge to the same level. This test is a one-sided t-test and the critical value at the 5% level is -1.65.
- Group composition and shares:
  - Group 2:
    - Largest group: 98 countries.
    - Accounts for 94 percent of global military expenditure on average.
    - Includes China, India, Russia, the United Kingdom, and the United States.
    - 30 of the 98 countries are advanced economies.
    - Average military outlays fell significantly from 1990 through the mid-2000s and have changed little since then.
  - Group 3:
    - 27 countries.
    - Accounts for 2.2 percent of total global military expenditure.
    - Only 2 advanced economies: Lithuania and Slovenia.
    - Converged to a lower level of about 1 percent of GDP.
  - Group 1:
    - 13 countries.
    - Smallest group but trends upward.
    - Accounts for 4.1 percent of global military expenditure.
    - Includes high-spending conflict countries such as Oman, Libya, and Saudi Arabia.
    - 2 of 13 are advanced economies.
- Group-level trajectories:
  - Group 2 and Group 3 follow the same downward trend after 1990 but at different rates.
  - For the largest group (about 90 percent of total global military spending in one description), outlays plateau at around 1.8 percent of GDP.
  - For the smaller high-conflict group of 13 countries, spending has risen to a substantially higher level and diverged from the global trend.
  - For the third group (27 countries), spending has converged to about 1 percent of GDP.

### VI. Determinants and implications highlighted
- Determinants examined include:
  - Global security environment (post–Cold War decline).
  - Fiscal pressures and high public debt in advanced economies (average debt-to-GDP ratio exceeding 100 percent in the post-financial crisis period).
  - Reallocation toward social spending and infrastructure in developing countries to meet SDG objectives.
  - Aging populations in advanced economies raising age-related spending pressures (pension and health).
- Main inferences:
  - Military spending in relation to GDP is converging but into three separate groups rather than a single global steady state.
  - Further declines in military spending are possible in developing economies, especially if conflict reduction and increases in social spending are realized.
  - In advanced economies, social spending does not appear to exert much downward pressure on military outlays (as stated in the text).
- Additional analyses to be reported in subsequent sections (as indicated in the text):
  - Determinants of military spending using an approach that captures the effect of social spending on military outlays.
  - Assessment of the impact of IMF-supported programs on military spending, revisiting prior work (Gupta, de Mello and Sharan (2001)).

### Appendix 1). Determinants of Military Spending — Overview and Baseline Specification
- Study period and focus: determinants of military spending over the past 47 years; aim to identify factors driving changes in military spending and implications for future outlays.
- Methodological approach: error correction model (following Kaufman and Segura-Ubiergo (2001)) separating short-term and long-term effects, assuming a long-term cointegrating relationship Y* = γ X*.
- Baseline error correction formulation:
  - ΔY_it = α + ΔX_i,t−1 + δ (Y_i,t−1 − γ X_i,t−1) + ε_it  (equation 10)
  - Transformed model: ΔY_it = α + δ Y_i,t−1 + Z_j ΔX_i,t−1 + β_j X_i,t−1 + ε_it  (equation 11)
- Definitions and interpretation:
  - ΔY_it: change in military expenditure of country i in year t, measured in percent of GDP.
  - Y_i,t−1: lagged level of military spending to GDP; coefficient δ measures annual correction rate toward long-term equilibrium.
  - X_i,t−1: vector of independent variables in levels (includes IMF program dummy IMF_it and social spending Social).
  - ΔX_it: first differences of these variables.
  - Long-term effect calculation: long-term effect of a permanently higher level of social spending captured by β_j divided by −δ (per Kaufman and Segura).
- Independent variable categories:
  1. IMF program presence: dummy = 1 if country has an IMF-supported program in that year, 0 otherwise.
  2. Security needs:
     - External threats: average neighbor’s military spending (simple average of military expenditures of border countries per CIA World Factbook).
     - Internal threats: political stability and absence of violence/terrorism from World Governance Indicators (WGI), split into "high" (above sample mean) and "low" (below sample mean).
  3. Governance: polity score (range −10 to 10; lower = total autocracy, upper = total democracy).
  4. Social spending: government health and education expenditure (World Bank); log of real GDP per capita (IMF WEO) included as control.

### Appendix 1). Empirical Results (1996–2014; country- and time-fixed effects)
- Estimation sample and notation:
  - Period covered: 1996 to 2014.
  - One-period lagged level variables designated “L.”; first differences designated “D.”.
- Key coefficient estimates from Table 4 (dependent variable: annual change in military spending as a share of GDP):
  - L. Military Expenditure (as percentage of GDP):
    - All: −0.3364***  [0.0427]
    - AEs: −0.1997***  [0.0322]
    - DEs: −0.3450***  [0.0455]
  - L. IMF Program:
    - All: 0.0087  [0.0396]
    - AEs: −0.0567  [0.0477]
    - DEs: 0.023  [0.0434]
  - D. IMF Program:
    - All: −0.0076  [0.0331]
    - AEs: 0.0229  [0.0300]
    - DEs: −0.0089  [0.0369]
  - L. GDP per Capita:
    - All: 0.1756  [0.2082]
    - AEs: −0.0361  [0.1722]
    - DEs: 0.1642  [0.2465]
  - D. GDP per Capita:
    - All: −0.2553  [0.5297]
    - AEs: 0.5977  [0.4227]
    - DEs: −0.3902  [0.6073]
  - L. Polity Score:
    - All: 0.0027  [0.0071]
    - AEs: 0.0046  [0.0161]
    - DEs: 0.0005  [0.0077]
  - D. Polity Score:
    - All: 0.0043  [0.0057]
    - AEs: −0.0428**  [0.0185]
    - DEs: 0.0055  [0.0061]
  - L. High Political Stability and Absence of Violence/Terrorism:
    - All: −0.0735  [0.0587]
    - AEs: −0.1637**  [0.0664]
    - DEs: −0.0774  [0.0638]
  - D. High Political Stability and Absence of Violence/Terrorism:
    - All: −0.1937  [0.1331]
    - AEs: −0.2284***  [0.0470]
    - DEs: −0.186  [0.1356]
  - L. Low Political Stability and Absence of Violence/Terrorism:
    - All: −0.1264  [0.0787]
    - AEs: −0.0227  [0.0684]
    - DEs: −0.1505*  [0.0900]
  - D. Low Political Stability and Absence of Violence/Terrorism:
    - All: 0.0162  [0.0588]
    - AEs: −0.0147  [0.0369]
    - DEs: 0.0235  [0.0708]
  - L. Average Neighbors' Spending (as percentage of GDP):
    - All: 0.0859***  [0.0242]
    - AEs: 0.0553  [0.0624]
    - DEs: 0.0855***  [0.0241]
  - D. Average Neighbors' Spending (as percentage of GDP):
    - All: −0.0977*  [0.0512]
    - AEs: 0.0512  [0.2034]
    - DEs: −0.1074**  [0.0518]
  - L. Health and Education Spending (as percentage of GDP):
    - All: −0.0379***  [0.0143]
    - AEs: −0.0054  [0.0094]
    - DEs: −0.0405***  [0.0146]
  - D. Health and Education Spending (as percentage of GDP):
    - All: 0.0092  [0.0195]
    - AEs: 0.0162  [0.0104]
    - DEs: 0.0061  [0.0217]
  - Observations (Obs.):
    - All: 1357
    - AEs: 317
    - DEs: 1040
  - R-Squared:
    - All: 0.31
    - AEs: 0.42
    - DEs: 0.32
  - Note: Country− and time-fixed effects included; robust standard errors clustered at country level. *, **, *** indicate significance at the 10, 5, and 1 percent level, respectively.
- Main empirical findings summarized:
  - Lagged military spending: negative and significant — military spending has fallen fastest for countries with highest initial levels of spending (relationship holds for both AEs and DEs; relatively stronger in DEs).
  - IMF programs: no statistically significant effect on military spending in short or long run.
  - Internal security concerns: important determinant—both levels and first differences significant and negative in advanced economies; significant long-term relationship in developing economies. Example: improving political stability and absence of violence/terrorism for countries from the worst 25 percentile of developing countries to the top 75 percent would, over the long run, reduce spending by about ½ a percent of GDP.
  - External threats (neighbors’ spending): positive long-term relationship in developing economies; example: a 1 percent of GDP decline in neighbors’ spending in developing economies would reduce a country’s outlays by 0.25 percent of GDP (based on long-term coefficient).
  - Governance (polity score): mixed results — higher polity score does not appear to affect defense outlays in developing countries; in advanced economies, higher polity score reduces military spending only in short run and coefficient is small.
  - Social spending: no crowding out in advanced economies; in developing economies, long-term relationship shows social spending crowds out defense. Example: an increase in social spending of 5 percentage points of GDP would cut defense spending by ½ percentage point of GDP over the long term.

### Appendix 1. Convergence Groups, 1970-2018
- Group 1:
  - Algeria, Armenia, Azerbaijan, Chad, Colombia, Ecuador, Estonia, Latvia, Libya, Oman, Saudi Arabia, Sudan, Ukraine.
- Group 2:
  - Angola, Argentina, Australia, Austria, Bahrain, Bangladesh, Belarus, Belgium, Belize, Benin, Bolivia, Botswana, Brazil, Brunei Darussalam, Burkina Faso, Burundi, Cambodia, Cameroon, Canada, China, Democratic Republic of the Congo, Republic of Congo, Croatia, Cyprus, Czech Republic, Côte d'Ivoire, Denmark, Dominican Republic, Fiji, Finland, France, Gabon, Gambia, Germany, Ghana, Greece, Guatemala, Hungary, India, Iran, Iraq, Ireland, Israel, Italy, Jamaica, Japan, Jordan, Korea, Kuwait, Lebanon, Lesotho, Luxembourg, Mali, Malta, Mauritania, Mauritius, Mexico, Moldova, Morocco, Myanmar, Namibia, Nepal, Netherlands, New Zealand, Niger, Norway, Pakistan, Papua New Guinea, Peru, Philippines, Poland, Portugal, Romania, Russia, Rwanda, Senegal, Sierra Leone, Singapore, Slovak Republic, South Africa, Spain, Sri Lanka, Swaziland, Sweden, Switzerland, Syria, Taiwan, Thailand, Togo, Trinidad and Tobago, Tunisia, Turkey, Uganda, United Kingdom, United States, Uruguay, Yemen, Zimbabwe.
- Group 3:
  - Albania, Bulgaria, Cabo Verde, Chile, Egypt, El Salvador, Ethiopia, Guyana, Honduras, Indonesia, Kenya, Liberia, Lithuania, Madagascar, Malawi, Malaysia, Mongolia, Mozambique, Nicaragua, Nigeria, Panama, Paraguay, Seychelles, Slovenia, Tanzania, Venezuela, Zambia.

### Appendix 2. Estimating Long-Term Effects — Key points and long-term coefficients
- Long-term effect computation: divide β (coefficients of lagged-level variables) by coefficient of lagged military spending times minus 1 (−δ) to obtain long-term impact on military spending as share of GDP.
- Summary statement: For whole sample, long-term trend in military spending is negatively associated with social spending and positively associated with neighbors’ military outlays; results driven by developing economies. In both AEs and DEs, high political stability and absence of violence/terrorism is negatively associated with military spending.
- Appendix 2.1. Long Term Coefficients for Determinants of Military Spending:
  - Health and Education Spending (as percentage of GDP):
    - All: −0.1127***
    - AEs: −0.0270
    - DEs: −0.1174***
  - Average Neighbors' Spending (as percentage of GDP):
    - All: 0.2554***
    - AEs: 0.2769
    - DEs: 0.2478***
  - Polity Score:
    - All: 0.0080
    - AEs: 0.0230
    - DEs: 0.0014
  - GDP per Capita:
    - All: 0.5220
    - AEs: −0.1808
    - DEs: 0.4759
  - IMF Program:
    - All: 0.0259
    - AEs: −0.2839
    - DEs: 0.0667
  - Low Political Stability and Absence of Violence/Terrorism:
    - All: −0.2185
    - AEs: −0.8197**
    - DEs: −0.2243
  - High Political Stability and Absence of Violence/Terrorism:
    - All: −0.3757
    - AEs: −0.1137
    - DEs: −0.4362*
- Note: Robust standard errors clustered at country level. *, **, and *** indicate significance at the 10, 5, and 1 percent level, respectively.

*Source: SIPRI, IMF World Economic Outlook, World Bank World Development Indicators, World Bank World Governance Indicators, and authors’ estimates (content excerpt from wpiea2019196-print-pdf - References).*

### References .............................................................................................................

### wpiea2019196-print-pdf - References

### I. Introduction and key trends
- Worldwide military spending (unweighted country averages) declined from 3.6 percent of GDP during the Cold War period (1970-1990) to 1.9 percent of GDP during 2010-2018.
- The share of the budget allocated to defense fell by three percentage points between the Cold War period and 2010-2018.
- Disaggregation:
  - Developing countries experienced a sharper fall in military spending as a share of total spending than advanced economies.
  - Total global military spending annual averages:
    - US$410,927 million during 1970-1990.
    - US$1,703,810 million during 2010-2018.
- Regional patterns:
  - Middle East average military spending: 4.5 percent of GDP (1992-2018 period).
  - Western Hemisphere average: around 1.4 percent of GDP.
  - Sub-Saharan Africa averaged 2.1 percent of GDP for the entire period.
  - Nearly half of Sub-Saharan African countries are classified as fragile states.
- Distribution of countries by military spending (1970-2018):
  - 57 countries with military spending under 2 percent of GDP.
  - 68 countries with military spending between 2-5 percent of GDP.
  - 13 countries with military spending in excess of 5 percent of GDP.
  - 100 countries allocate less than 10 percent of the budget to the military.
  - 31 countries allocate between 10 and 20 percent of the budget to the military.
  - 7 countries allocate more than 20 percent of the budget to the military.
- Several advanced economies are among the top 15 military spenders and account for over 80 percent of worldwide military spending.

### II. Summary statistics (Table 1)
- Military expenditure as percent of GDP — means and standard deviations by period and group:
  - All:
    - Mean: 1970-1990 = 3.61; 1992-2000 = 2.79; 2001-2009 = 2.10; 2010-2018 = 1.90
    - St.Dev.: (3.79) (3.09) (1.80) (1.60)
  - Advanced:
    - Mean: 1970-1990 = 3.60; 1992-2000 = 2.23; 2001-2009 = 1.85; 2010-2018 = 1.59
    - St.Dev.: (3.80) (1.72) (1.29) (1.01)
  - Developing:
    - Mean: 1970-1990 = 3.62; 1992-2000 = 2.98; 2001-2009 = 2.18; 2010-2018 = 1.99
    - St.Dev.: (3.72) (3.40) (1.93) (1.73)
  - Sub-Saharan Africa:
    - Mean: 1970-1990 = 2.61; 1992-2000 = 2.78; 2001-2009 = 1.87; 2010-2018 = 1.67
    - St.Dev.: (2.27) (4.44) (2.18) (1.27)
- Military expenditure as percent of total expenditure — means by period (All):
  - Mean: 1970-1990 = 9.58; 1992-2000 = 9.12; 2001-2009 = 7.46; 2010-2018 = 6.42
  - Advanced means: 6.98, 5.31, 5.07, 4.33 (1970-1990; 1992-2000; 2001-2009; 2010-2018)
  - Developing means: 12.58, 10.54, 8.20, 7.07 (same periods)
- Sources: SIPRI, IMF World Economic Outlook.

### III. Explanatory factors cited in the literature
- Factors proposed to explain the decline in military spending:
  - End of the Cold War and improved global security environment.
  - Fiscal consolidation pressures in advanced economies, with average debt-to-GDP ratio exceeding 100 percent in the post-financial crisis period (IMF 2018).
  - Developing countries shifting budget shares toward social sectors and infrastructure since the early 2000s, driven by growth and SDG needs.
  - Aging populations in advanced economies increasing pension and health spending pressures (Clements et al. 2017).
- Prior literature and methods referenced:
  - AKA (Gleditsch et al. (1996), Gupta, Schiff, and Clements (1998), Rockoff (1998), Avranitidis, Kollias and Anestasopoulos (2014)) studied sigma and beta convergence for 128 countries during 1989-2008.
  - Lau, Demir, and Bilgin (2015) used nonlinear unit root tests on 37 countries (1988-2012) and analyzed pairwise convergence thresholds.

### IV. Convergence testing: data and methods
- Data:
  - SIPRI annual military expenditure data for 140 economies (34 advanced economies, 106 developing economies) covering 1970-2017.
  - SIPRI includes current and capital spending on armed and paramilitary forces, government agencies engaged in defense projects, and military space activities; expenditures on personnel (including pensions), operations and maintenance, procurement and construction, research and development, civil defense, and demobilization.
- Tests applied:
  - Sigma convergence (cross-sectional variance over time) shows convergence for advanced economies, developing countries, and the world as a whole (Figure 4).
  - Cross-sectional beta convergence baseline specification (dependent variable: average annual change in military expenditure as percent of GDP over T; independent variable: initial y).
    - Table 2 results (Cross-Sectional Beta Convergence Test):
      - Coefficient on y_0:
        - All: -0.0237*** [0.0018]
        - Advanced: -0.0250*** [0.0012]
        - DE: -0.0236*** [0.0023]
      - Cons:
        - All: 0.0364*** [0.0082]
        - Advanced: 0.0277*** [0.0051]
        - DE: 0.0400*** [0.0106]
      - Obs.: 140 (All); 34 (Advanced); 106 (DE)
      - R-Sq: 0.56 (All); 0.93 (Advanced); 0.51 (DE)
    - Interpretation: For the whole sample the point estimate of the beta coefficient is -0.02; a one-percentage higher initial value is associated with a 0.2 percentage faster decrease in military expenditure as a percentage of GDP.
  - Phillips and Sul (2007) club convergence methodology:
    - Decomposes y_it = phi_i * mu_t + eps_it and constructs relative transition coefficients h_it and cross-sectional variance V_t.
    - Tests convergence by estimating log(V_1) - log(V_t) - 2 log L(t) = a + b log t + eps_t for t = rT,...,T and testing whether b >= 0 (equivalently alpha >= 0).
    - Practical modeling uses delta_it = delta_i + sigma_i * theta_it * L(t) * t^(-alpha).

### V. Convergence group classification and patterns (Table 3, Figures 5–6)
- Convergence group test results (Table 3):
  - Group 1:
    - Countries: 13
    - Time periods: 49
    - Coefficient: 0.55**
    - S.E.: 0.35
    - t-stat.: 1.58
  - Group 2:
    - Countries: 98
    - Time periods: 49
    - Coefficient: -0.02**
    - S.E.: 0.01
    - t-stat.: -1.60
  - Group 3:
    - Countries: 27
    - Time periods: 49
    - Coefficient: -0.19**
    - S.E.: 0.23
    - t-stat.: -0.82
  - Note: The null hypothesis is that all countries within each group converge to the same level. This test is a one-sided t-test and the critical value at the 5% level is -1.65.
- Group composition and shares:
  - Group 2:
    - Largest group: 98 countries.
    - Accounts for 94 percent of global military expenditure on average.
    - Includes China, India, Russia, the United Kingdom, and the United States.
    - 30 of the 98 countries are advanced economies.
    - Average military outlays fell significantly from 1990 through the mid-2000s and have changed little since then.
  - Group 3:
    - 27 countries.
    - Accounts for 2.2 percent of total global military expenditure.
    - Only 2 advanced economies: Lithuania and Slovenia.
    - Converged to a lower level of about 1 percent of GDP (described elsewhere in the text).
  - Group 1:
    - 13 countries.
    - Smallest group but trends upward.
    - Accounts for 4.1 percent of global military expenditure.
    - Includes high-spending conflict countries such as Oman, Libya, and Saudi Arabia.
    - 2 of 13 are advanced economies.
- Group-level trajectories:
  - Group 2 and Group 3 follow the same downward trend after 1990 but at different rates.
  - For the largest group (about 90 percent of total global military spending in one description), outlays plateau at around 1.8 percent of GDP.
  - For the smaller high-conflict group of 13 countries, spending has risen to a substantially higher level and diverged from the global trend.
  - For the third group (27 countries), spending has converged to about 1 percent of GDP.

### VI. Determinants and implications highlighted
- Determinants examined include:
  - Global security environment (post–Cold War decline).
  - Fiscal pressures and high public debt in advanced economies (average debt-to-GDP ratio exceeding 100 percent in the post-financial crisis period).
  - Reallocation toward social spending and infrastructure in developing countries to meet SDG objectives.
  - Aging populations in advanced economies raising age-related spending pressures (pension and health).
- Main inferences:
  - Military spending in relation to GDP is converging but into three separate groups rather than a single global steady state.
  - Further declines in military spending are possible in developing economies, especially if conflict reduction and increases in social spending are realized.
  - In advanced economies, social spending does not appear to exert much downward pressure on military outlays (as stated in the text).
- Additional analyses to be reported in subsequent sections (as indicated in the text):
  - Determinants of military spending using an approach that captures the effect of social spending on military outlays.
  - Assessment of the impact of IMF-supported programs on military spending, revisiting prior work (Gupta, de Mello and Sharan (2001)).

*Source: SIPRI, IMF World Economic Outlook, and authors’ calculations (content excerpt from wpiea2019196-print-pdf - References).*

### Appendix 1).

### Appendix 1)

### Determinants of Military Spending — Overview
- Study period and focus: determinants of military spending over the past 47 years; aim to identify factors driving changes in military spending and implications for future outlays.
- Methodological approach: error correction model (following Kaufman and Segura-Ubiergo (2001)) separating short-term and long-term effects, assuming a long-term cointegrating relationship Y* = γ X*.

### Baseline Specification
- Error correction formulation (as presented):
  - ΔY_it = α + ΔX_i,t−1 + δ (Y_i,t−1 − γ X_i,t−1) + ε_it  (equation 10)
  - Transformed model: ΔY_it = α + δ Y_i,t−1 + Z_j ΔX_i,t−1 + β_j X_i,t−1 + ε_it  (equation 11)
  - Definitions:
    - ΔY_it: change in military expenditure of country i in year t, measured in percent of GDP.
    - Y_i,t−1: lagged level of military spending to GDP; coefficient δ measures annual correction rate toward long-term equilibrium.
    - X_i,t−1: vector of independent variables in levels (includes IMF program dummy IMF_it and social spending Social).
    - ΔX_it: first differences of these variables.
  - Long-term effect calculation: long-term effect of a permanently higher level of social spending captured by β_j divided by −δ (per Kaufman and Segura).

- Independent variable categories:
  1. IMF program presence: dummy = 1 if country has an IMF-supported program in that year, 0 otherwise.
  2. Security needs:
     - External threats: average neighbor’s military spending (simple average of military expenditures of border countries per CIA World Factbook).
     - Internal threats: political stability and absence of violence/terrorism from World Governance Indicators (WGI), split into "high" (above sample mean) and "low" (below sample mean).
  3. Governance: polity score (range −10 to 10; lower = total autocracy, upper = total democracy).
  4. Social spending: government health and education expenditure (World Bank); log of real GDP per capita (IMF WEO) included as control.

### Empirical Results (1996–2014; country- and time-fixed effects)
- Estimation sample and notation:
  - Period covered: 1996 to 2014 (due to data limitations).
  - One-period lagged level variables designated “L.”; first differences designated “D.”.
- Key coefficient estimates from Table 4 (dependent variable: annual change in military spending as a share of GDP):
  - L. Military Expenditure (as percentage of GDP):
    - All: −0.3364***  [0.0427]
    - AEs: −0.1997***  [0.0322]
    - DEs: −0.3450***  [0.0455]
  - L. IMF Program:
    - All: 0.0087  [0.0396]
    - AEs: −0.0567  [0.0477]
    - DEs: 0.023  [0.0434]
  - D. IMF Program:
    - All: −0.0076  [0.0331]
    - AEs: 0.0229  [0.0300]
    - DEs: −0.0089  [0.0369]
  - L. GDP per Capita:
    - All: 0.1756  [0.2082]
    - AEs: −0.0361  [0.1722]
    - DEs: 0.1642  [0.2465]
  - D. GDP per Capita:
    - All: −0.2553  [0.5297]
    - AEs: 0.5977  [0.4227]
    - DEs: −0.3902  [0.6073]
  - L. Polity Score:
    - All: 0.0027  [0.0071]
    - AEs: 0.0046  [0.0161]
    - DEs: 0.0005  [0.0077]
  - D. Polity Score:
    - All: 0.0043  [0.0057]
    - AEs: −0.0428**  [0.0185]
    - DEs: 0.0055  [0.0061]
  - L. High Political Stability and Absence of Violence/Terrorism:
    - All: −0.0735  [0.0587]
    - AEs: −0.1637**  [0.0664]
    - DEs: −0.0774  [0.0638]
  - D. High Political Stability and Absence of Violence/Terrorism:
    - All: −0.1937  [0.1331]
    - AEs: −0.2284***  [0.0470]
    - DEs: −0.186  [0.1356]
  - L. Low Political Stability and Absence of Violence/Terrorism:
    - All: −0.1264  [0.0787]
    - AEs: −0.0227  [0.0684]
    - DEs: −0.1505*  [0.0900]
  - D. Low Political Stability and Absence of Violence/Terrorism:
    - All: 0.0162  [0.0588]
    - AEs: −0.0147  [0.0369]
    - DEs: 0.0235  [0.0708]
  - L. Average Neighbors' Spending (as percentage of GDP):
    - All: 0.0859***  [0.0242]
    - AEs: 0.0553  [0.0624]
    - DEs: 0.0855***  [0.0241]
  - D. Average Neighbors' Spending (as percentage of GDP):
    - All: −0.0977*  [0.0512]
    - AEs: 0.0512  [0.2034]
    - DEs: −0.1074**  [0.0518]
  - L. Health and Education Spending (as percentage of GDP):
    - All: −0.0379***  [0.0143]
    - AEs: −0.0054  [0.0094]
    - DEs: −0.0405***  [0.0146]
  - D. Health and Education Spending (as percentage of GDP):
    - All: 0.0092  [0.0195]
    - AEs: 0.0162  [0.0104]
    - DEs: 0.0061  [0.0217]
  - Observations (Obs.):
    - All: 1357
    - AEs: 317
    - DEs: 1040
  - R-Squared:
    - All: 0.31
    - AEs: 0.42
    - DEs: 0.32
  - Note: Country− and time-fixed effects included; robust standard errors clustered at country level. *, **, *** indicate significance at the 10, 5, and 1 percent level, respectively.

- Main empirical findings summarized in text:
  - Lagged military spending: negative and significant — military spending has fallen fastest for countries with highest initial levels of spending (relationship holds for both AEs and DEs; relatively stronger in DEs).
  - IMF programs: no statistically significant effect on military spending in short or long run.
  - Internal security concerns: important determinant—both levels and first differences significant and negative in advanced economies; significant long-term relationship in developing economies. Example: improving political stability and absence of violence/terrorism for countries from the worst 25 percentile of developing countries to the top 75 percent would, over the long run, reduce spending by about ½ a percent of GDP.
  - External threats (neighbors’ spending): positive long-term relationship in developing economies; example from text and Appendix 2: a 1 percent of GDP decline in neighbors’ spending in developing economies would reduce a country’s outlays by 0.25 percent of GDP (based on long-term coefficient).
  - Governance (polity score): mixed results — higher polity score does not appear to affect defense outlays in developing countries; in advanced economies, higher polity score reduces military spending only in short run and coefficient is small.
  - Social spending: no crowding out in advanced economies; in developing economies, long-term relationship shows social spending crowds out defense. Example: an increase in social spending of 5 percentage points of GDP would cut defense spending by ½ percentage point of GDP over the long term.

### Concluding Remarks (summary points from text)
- Convergence: using Phillips and Sul technique, convergence into 3 distinct groups:
  - Group 2 (largest) accounting for over 90 percent of global military spending: outlays fell from Cold War levels and plateaued at an average of around 1.8 percent of GDP.
  - Small group (conflict-driven): outlays increased to about 7 percent of GDP.
  - Primarily developed economies group: spending fell from around 3 percent of GDP pre–Cold War to about 1 percent of GDP.
- Determinants recap:
  - No significant effect from IMF programs (unlike earlier studies).
  - No impact of governance on military spending in developing economies found.
  - External threat levels matter in developing economies (neighbors’ spending).
  - Reducing internal conflict yields a substantial “peace dividend”; for DEs, improvements in internal threat measures from bottom 25 percent to top 25 percent could reduce military spending by ½ percent of GDP.
  - Rising health and education spending in DEs (to meet SDGs) would likely reduce military spending.

### Appendix 1. Convergence Groups, 1970-2018
- Group 1:
  - Algeria, Armenia, Azerbaijan, Chad, Colombia, Ecuador, Estonia, Latvia, Libya, Oman, Saudi Arabia, Sudan, Ukraine.
- Group 2:
  - Angola, Argentina, Australia, Austria, Bahrain, Bangladesh, Belarus, Belgium, Belize, Benin, Bolivia, Botswana, Brazil, Brunei Darussalam, Burkina Faso, Burundi, Cambodia, Cameroon, Canada, China, Democratic Republic of the Congo, Republic of Congo, Croatia, Cyprus, Czech Republic, Côte d'Ivoire, Denmark, Dominican Republic, Fiji, Finland, France, Gabon, Gambia, Germany, Ghana, Greece, Guatemala, Hungary, India, Iran, Iraq, Ireland, Israel, Italy, Jamaica, Japan, Jordan, Korea, Kuwait, Lebanon, Lesotho, Luxembourg, Mali, Malta, Mauritania, Mauritius, Mexico, Moldova, Morocco, Myanmar, Namibia, Nepal, Netherlands, New Zealand, Niger, Norway, Pakistan, Papua New Guinea, Peru, Philippines, Poland, Portugal, Romania, Russia, Rwanda, Senegal, Sierra Leone, Singapore, Slovak Republic, South Africa, Spain, Sri Lanka, Swaziland, Sweden, Switzerland, Syria, Taiwan, Thailand, Togo, Trinidad and Tobago, Tunisia, Turkey, Uganda, United Kingdom, United States, Uruguay, Yemen, Zimbabwe.
- Group 3:
  - Albania, Bulgaria, Cabo Verde, Chile, Egypt, El Salvador, Ethiopia, Guyana, Honduras, Indonesia, Kenya, Liberia, Lithuania, Madagascar, Malawi, Malaysia, Mongolia, Mozambique, Nicaragua, Nigeria, Panama, Paraguay, Seychelles, Slovenia, Tanzania, Venezuela, Zambia.

### Appendix 2. Estimating Long-Term Effects — Key points and long-term coefficients
- Long-term effect computation: divide β (coefficients of lagged-level variables) by coefficient of lagged military spending times minus 1 (−δ) to obtain long-term impact on military spending as share of GDP.
- Summary statement: For whole sample, long-term trend in military spending is negatively associated with social spending and positively associated with neighbors’ military outlays; results driven by developing economies. In both AEs and DEs, high political stability and absence of violence/terrorism is negatively associated with military spending.

- Appendix 2.1. Long Term Coefficients for Determinants of Military Spending:
  - Health and Education Spending (as percentage of GDP):
    - All: −0.1127***
    - AEs: −0.0270
    - DEs: −0.1174***
  - Average Neighbors' Spending (as percentage of GDP):
    - All: 0.2554***
    - AEs: 0.2769
    - DEs: 0.2478***
  - Polity Score:
    - All: 0.0080
    - AEs: 0.0230
    - DEs: 0.0014
  - GDP per Capita:
    - All: 0.5220
    - AEs: −0.1808
    - DEs: 0.4759
  - IMF Program:
    - All: 0.0259
    - AEs: −0.2839
    - DEs: 0.0667
  - Low Political Stability and Absence of Violence/Terrorism:
    - All: −0.2185
    - AEs: −0.8197**
    - DEs: −0.2243
  - High Political Stability and Absence of Violence/Terrorism:
    - All: −0.3757
    - AEs: −0.1137
    - DEs: −0.4362*

- Note: Robust standard errors clustered at country level. *, **, and *** indicate significance at the 10, 5, and 1 percent level, respectively.

*Source: SIPRI, IMF World Economic Outlook, World Bank World Development Indicators, World Bank World Governance Indicators, and authors’ estimates (as presented in the source PDF).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019196-print-pdf.pdf_