## Macroeconomic Impacts of EU Defense Spending

## Source details

**Canonical URL:** [Macroeconomic Impacts of EU Defense Spending](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026053-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026053-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026053-source-pdf.pdf.json)

---

### I. Key findings and context
- Between 2021 and 2024, defense outlays among European Union (EU) members increased by about 30 percent, reaching 1.9 of the EU’s GDP in 2024.
- All NATO allies are expected to meet or exceed the 2 percent of GDP spending target (NATO, August 2025).
- Proposals under discussion aim for defense and security related spending to reach 5 percent of GDP by 2035.
- Historical analysis (1989–2023, 27 EU countries) indicates:
  - Estimated short-term defense spending multipliers are statistically significant and above one.
  - Multipliers become less precise at longer horizons.
  - Multipliers align with recent EU estimates (Garcia-Serrador et al. 2025; Ben Zeev et al. 2025) and US state-level evidence (Nakamura and Steinsson 2014).

### II. Research questions, data, and empirical approach
- Core questions:
  - Macroeconomic effects of past increases in defense spending in EU countries.
  - Transmission channels and state-dependent variation of defense multipliers.
  - Cross-border spillovers and anticipation effects from procurement.
  - Heterogeneous impacts by procurement category (construction, services, equipment, R&D).
- Data and methods:
  - Country-level panel: annual defense spending outlays for 27 EU countries, 1989–2023, using local projections per Ramey and Zubairy (2018).
  - High-frequency dataset: monthly defense procurement obligations (award notices) for EU-27, 2009–2023.
  - Identification strategies: OLS local projections; instrumental variables using aggregate European defense spending interacted with country-specific geographic exposure (distance to Russia; land-border dummy with selected non-EU, non-NATO neighbors).
  - Procurement treated as fiscal news that typically precedes payments by about three to four quarters.

### III. Aggregate effects and multipliers (annual outlays)
- Output and multiplier responses to a 1 percent of trend GDP increase in defense spending:
  - Output rises by "about 1.3 percent" in the same year and peaks at "1.6 percent" one year after the shock; output response is significant for two years.
  - Baseline cumulative multiplier is "approximately 1.3 on impact" and rises to "about 1.9 two years after the shocks"; baseline cumulative multiplier peaks "around 2".
  - Aggregate IV estimate: roughly 1.9 at 36 months.
  - Estimated defense-spending multipliers reported: 1.5 at the 12-month horizon; approximately 1.7 at two years; around 1.9 after three years.
- Demand and supply components (response to a 1 percent of trend GDP shock):
  - Government consumption: increased by about 0.5 percent of GDP in the first year and government consumption increases by about 0.5 percent of trend GDP in response to a defense spending shock.
  - Private consumption: rises by about 1 percent of trend GDP (crowd-in).
  - Investment: increases by about 1 percent of trend GDP and displays greater persistence than consumption.
  - Imports: net imports rise; a 1 percent of trend GDP shock raises total imports by about 2 percent of GDP (with intra-EU and extra-EU contributions roughly equal).
- Supply-side and prices:
  - Capital and labor services expand; TFP increases are confined to cyclical components and do not extend to trend TFP.
  - Consumer prices increase, implying positive demand effects exceed positive supply effects.
- Interest rates and exchange rates:
  - Short-term nominal interest rates tended to decline following defense shocks in the sample period.
  - Real exchange rates depreciated following defense shocks; high-frequency evidence shows REER depreciations of about 0.4 percent at two years and 0.7 percent at three years.
  - A 1 percent increase in defense spending leads to a cumulative real depreciation of the effective exchange rate of roughly 4 percent relative to trend at the two- and three-year horizons (reported in one section).

### IV. Heterogeneity of multipliers and transmission channels
- Multipliers are systematically larger when:
  - Import intensity is low (output multiplier one year after shock ≈ 3 in low–import-intensity environments vs. ≈ 1 in high–import-intensity environments).
  - Fiscal space is ample (low sovereign spreads).
  - Public investment efficiency is high (multiplier of 3 in the first three years under high efficiency vs. 1 under low efficiency).
- Little evidence of systematic asymmetry between positive and negative defense spending shocks.
- Procurement-category heterogeneity (OLS and IV):
  - Equipment-biased packages: largest medium-term multipliers, reaching about 2.7 at 36 months.
  - Services packages: multipliers around 1 after 24 months and 1.2 after three years.
  - Construction-biased packages: multipliers near 0.6 at 36 months.
  - R&D-related packages: cumulative multiplier near 1.0 at the 36-month horizon (statistically significant at the 68% level).
  - IV estimates for subcomponents are similar to OLS and generally larger than OLS, with equipment spending standing out.

### V. Cross-border spillovers via trade
- Imports and arms imports:
  - A 1 percent of trend GDP defense spending shock raises total imports by about 2 percent of trend GDP.
    - Intra-EU imports increase by roughly 1 percent of trend GDP.
    - Extra-EU imports increase by roughly 1 percent of trend GDP.
  - Arms imports (SIPRI TIV): increase by 50 percent following a 1 percent of GDP defense shock, with responses from both intra-EU and extra-EU sources.
- Output spillovers:
  - A trade-weighted foreign defense spending shock S: a 1 percent of GDP increase in partner-country defense spending raises domestic GDP by about 0.4 percent after one year (0.2 percent on impact), corresponding to about 15 to 25 percent of the estimated domestic effect.
  - In coordinated increases, some spillovers could be offsetting.

### VI. High-frequency procurement evidence (monthly award notices)
- Advantages of procurement data:
  - Monthly award notices constitute fiscal news with timing plausibly unrelated to business-cycle fluctuations and precede disbursements by about three to four quarters.
  - Granularity permits category-specific analysis (construction, services, equipment, R&D).
- OLS procurement results:
  - Cumulative multipliers remain statistically significant after 12 months.
  - Multipliers reach approximately 0.7 after two years (24 months) and 0.8 after three years (36 months).
  - REER depreciations of about 0.4 percent at two years and 0.7 percent at three years.
- IV procurement results (shift–share instrument using aggregate EU procurement and country exposure):
  - IV estimates produce larger multipliers; cumulative multiplier reaches 1.5 at 12 months and more than twice OLS at many horizons.
  - First-stage diagnostics: Kleibergen–Paap rk Wald F statistic consistently exceeds 10 across horizons; Hansen J not rejected.
  - IV cumulative fiscal multipliers obtained are broadly consistent with annual IV estimates (e.g., ~1.9 at 36 months).
- Data construction and exclusions:
  - Sample period: 2009:m1 to 2023:m12 (EU-27).
  - Trend GDP estimated with a fourth-degree polynomial on log real GDP for 1995:m1-2023:12 excluding 2020:m3-2021:m12.
  - Outlier months dropped if defense procurement or total procurement exceeds 5 percent of annual GDP, monthly defense procurement > 12.5 percent of trend RGDP, or monthly total procurement > 50 percent of trend RGDP.

### VII. Identification, robustness, and diagnostics
- Annual outlays: local projection framework regresses cumulative output (normalized by trend GDP) on cumulative defense spending (normalized by trend GDP), controlling for macro variables; coefficients interpreted as fiscal multipliers under exogeneity assumptions.
- Monthly procurement IV: shift–share instrument constructed as EU-27 aggregate procurement normalized by EU trend GDP interacted with predetermined country exposure (population-weighted distance to Russia; land-border dummy with selected non-EU, non-NATO neighbors).
- Instrument diagnostics and robustness:
  - First-stage Cragg-Donald and Kleibergen–Paap statistics reported across horizons; Kleibergen–Paap rk Wald F > 10 for many horizons.
  - Hansen J p-values indicate overidentification not rejected across horizons.
  - Robustness checks: removing time fixed effects, controlling for revenue and public debt, alternative lag structures (1, 2, 3 lags), country-specific time trends, leave-one-out by country, pre-COVID and alternate sample windows. Results are broadly robust.
  - CLR (Finlay and Magnusson, 2009) weak-instrument-robust inference widens CIs but remains consistent with baseline.

### VIII. Policy implications, caveats, and suggested research avenues
- Policy-relevant implications:
  - Increased defense outlays generate positive output effects with multipliers often exceeding one, induce real exchange-rate depreciation, and produce meaningful intra-EU and extra-EU import responses.
  - Procurement composition matters: equipment-intensive packages yield the largest multipliers; construction and services produce positive but smaller multipliers; R&D shows moderate effects.
  - Structural characteristics (import intensity, fiscal space, public investment efficiency) materially affect multiplier size and implied fiscal outcomes.
- Caveats:
  - The ongoing large and simultaneous increase in defense spending across European countries differs from the historical 1989–2023 sample; not all historical estimates may translate directly to the current coordinated buildup.
  - Multipliers relevant to the current defense buildup might be lower than historical estimates if monetary policy becomes less accommodative going forward.
  - Anticipation and implementation lags can bias realized outlay-based multipliers downward; procurement data mitigate but do not eliminate all identification concerns.
- Suggested future research:
  - Alternative classifications of defense spending (automated contract parsing, keyword techniques).
  - Allocating expenditures over contract durations instead of full award-date allocation to separate anticipated from unanticipated shocks.
  - Incorporating supplier-level and place-of-performance data to map geographic and sectoral spillovers.
  - Jointly identifying shocks to defense and non-defense spending to compare multipliers.

*Italic source: IMF Working Paper — Macroeconomic Impacts of EU Defense Spending (wpiea2026053-source-pdf).*

### REFERENCES _____________________________________________________________________________________ 31

### Macroeconomic Impacts of EU Defense Spending

### I. Key findings and context
- Between 2021 and 2024, defense outlays among European Union (EU) members increased by about 30 percent, reaching 1.9 of the EU’s GDP in 2024.
- All NATO allies are expected to meet or exceed the 2 percent of GDP spending target (NATO, August 2025).
- Proposals under discussion aim for defense and security related spending to reach 5 percent of GDP by 2035.
- Historical analysis (1989–2023, 27 EU countries) indicates:
  - Estimated short-term defense spending multipliers are statistically significant and above one.
  - Multipliers become less precise at longer horizons.
  - Multipliers align with recent EU estimates (Garcia-Serrador et al. 2025; Ben Zeev et al. 2025) and US state-level evidence (Nakamura and Steinsson 2014).

### II. Research questions and scope
- Core questions addressed:
  - Macroeconomic effects of past increases in defense spending in EU countries.
  - Transmission channels for those effects.
  - Variation of defense multipliers with economic conditions and structural characteristics.
  - Cross-border spillovers from defense spending.
  - Anticipation effects revealed by high-frequency procurement awards.
  - Heterogeneous impacts by type of defense procurement.
- Data and approaches:
  - Country-level panel: annual defense spending outlays for 27 EU countries, 1989–2023, using local projections per Ramey and Zubairy (2018).
  - High-frequency dataset: monthly defense procurement obligations (award notices) for EU-27, 2009–2023, with instrumental variables using aggregate European defense spending and country-specific geographic distance to major adversaries.

### III. Aggregate effects and multipliers (annual outlays)
- A 1 percent of trend GDP increase in defense spending:
  - Raised government consumption by about 0.5 percent of GDP in the first year.
  - Increased investment and private consumption, signaling expansion of domestic absorption consistent with multipliers above one.
- Supply-side responses:
  - Positive responses in capital and labor services.
  - Effects on total factor productivity are confined to cyclical components and do not extend to trend TFP.
- Prices:
  - Consumer prices increase, implying positive demand effects exceed positive supply effects.
- Interest rates and exchange rates:
  - Short-term nominal interest rates tended to decline following defense shocks in the sample period.
  - Real exchange rates depreciated following defense shocks.
  - Patterns consistent with countries operating within a monetary union and with often accommodative monetary policy (policy rates at very low levels from 2009 through 2022).

### IV. Heterogeneity of multipliers by country characteristics and shock sign
- Multipliers are systematically larger when:
  - Import intensity is low.
  - Fiscal space is ample (proxied by sovereign yield spreads).
  - Public investment efficiency is high.
- Little evidence of systematic asymmetry between positive and negative defense spending shocks.

### V. Cross-border spillovers
- A 1 percent GDP increase in domestic defense spending raised imports by about 2 percent of GDP, with roughly equal contributions from intra-EU and extra-EU trade.
- Constructed “spillover shock” (export-share-weighted average of trading partners’ defense spending changes) findings:
  - A 1 percent of GDP increase in partner-country defense spending raised domestic GDP by about 0.4 percent after one year.
  - This corresponds to about one-quarter of the estimated domestic effect.
- In the case of coordinated increases across countries, some spillovers could be offsetting.

### VI. High-frequency procurement evidence (monthly award notices)
- Advantages of procurement data:
  - Higher frequency improves identification because contract timing is plausibly unrelated to business-cycle fluctuations.
  - Procurement records capture fiscal news (contractual obligations) that private agents respond to, preceding realized disbursements by about three to four quarters (Briganti et al., 2025).
  - Granularity enables category-specific analysis: construction, services, equipment, and research and development (R&D).
- OLS procurement results:
  - Cumulative multipliers remain statistically significant after 12 months.
  - Multipliers reach approximately 0.7 after two years and 0.8 after three years.
  - Real effective exchange rate (REER) depreciations of about 0.4 percent at two years and 0.7 percent at three years.
- Instrumental variable (IV) procurement results:
  - IV estimates produce larger multipliers, with the cumulative multiplier reaching 1.5 at 12 months, broadly matching annual-data results.
- Heterogeneity across procurement categories:
  - Spending is classified into construction, services, equipment, and R&D per Directive 2009/81.
  - Equipment spending generates the largest multipliers (detailed magnitudes and category-specific results are part of the disaggregated analysis).

### VII. Identification, robustness, and methodological notes
- Annual outlays analysis: local projection framework regresses cumulative output (normalized by trend GDP) on cumulative defense spending (normalized by trend GDP), controlling for common macroeconomic variables; coefficient on cumulative defense spending interpreted as the fiscal multiplier under exogeneity assumptions.
- Monthly procurement analysis:
  - Uses instrumental variables combining aggregate European defense spending with country-specific geographic distance to major adversaries derived from military alliances and treaty networks to address endogeneity (e.g., geopolitical risk correlated with both economic activity and defense spending).
  - Procurement obligations are treated as fiscal news that can be anticipated by private agents; payments typically realized after a lag of three to four quarters.
- Anticipation, measurement, and implementation biases in realized outlays can depress multiplier estimates; procurement data mitigate these biases.

### VIII. Caveats and implications for current/future defense buildup
- The ongoing large and simultaneous increase in defense spending across European countries differs from smaller, national historical efforts; not all historical estimates may translate directly to the current buildup.
- Multipliers relevant to the current defense buildup might be lower than historical estimates if monetary policy becomes less accommodative going forward.
- Cross-border effects imply coordinated or near-simultaneous increases across countries have important import and spillover considerations.

*Source: IMF Working Paper — Macroeconomic Impacts of EU Defense Spending (excerpted content).*

### 2.7 at 36 months, while services, construction, and R&D components also produce positive and statistically

### Macroeconomic Impacts of EU Defense Spending

### Key findings
- A defense spending shock of "1 percent of trend GDP" leads to:
  - an increase in defense spending that fades gradually (Figure 2, Panel A).
  - output rising by "about 1.3 percent" in the same year as the shock and reaching a peak of "1.6 percent" one year after the shock; output response is significant for two years (Figure 2, Panel B).
- The cumulative defense-spending multiplier:
  - is "approximately 1.3 on impact" and rises to "about 1.9 two years after the shocks" (Figure 2, Panel C).
  - baseline cumulative multiplier peaks "around 2" (statement comparing to related studies).
- Short-term interest rates and real exchange rates respond to defense shocks with:
  - short-term interest rates declining and the real exchange rate depreciating, or at a minimum no tightening of the monetary policy stance over the considered sample period (Figure 3, Panels A and B).
- Exogeneity evidence:
  - the war indicator is the only statistically significant predictor of defense spending in lagged-macroeconomic predictive regressions; coefficients on lagged output and government consumption are small and statistically insignificant.
  - defense spending shows no statistically significant relationship with a PCA-based anticipation measure constructed from October WEO projections of output growth, inflation, and government expenditure.
  - defense spending shocks are not significantly associated with monetary policy shocks or geopolitical risk shocks once controls are included.
- Sample and data:
  - unbalanced panel of the "27 EU countries" during "1989-2023".
  - GDP and government consumption from the United Nation’s National Accounts Aggregates Database; military spending from SIPRI; armed conflict index from UCDP/PRIO Armed Conflict Dataset and Miyamoto et al. (2019).

### Relation to existing literature
- Most existing studies use military spending as an instrument to identify government spending shocks (Ramey & Shapiro,1998; Ramey, 2011, 2016; Barro & Redlick, 2011; Nakamura & Steinsson, 2014; Antolin-Diaz & Surico, 2025).
- Comparisons to recent EU-focused estimates:
  - Ilzetzki (2025): Europe-wide GDP growth will increase by "0.9 percentage point (reaching about 1.5%)" if defense spending rises from "2% to 3.5% of GDP", implying a multiplier of "about 0.6 to 1".
  - Garcia-Serrador et al. (2025): a "1 percent increase in trend-GDP defense outlays raises aggregate output by about 1.4 percent within one year, peaking at 1.6 percent after two years", with the multiplier more pronounced in deep downturns "(exceeding 1.8)" compared to expansions "(about 0.8)".
  - Ben Zeev et al. (2025): output multipliers "around 2" for Europe and NATO members (three times larger than in the US), driven by higher R&D spending and TFP growth.
- Explanations for larger EU multipliers highlighted in the literature:
  - currency union or effectively fixed-exchange-rate regimes can amplify fiscal multipliers (Ilzetzki, Mendoza, and Végh, 2013; Cacciatore et al., 2021).
  - fiscal multipliers tend to be larger under accommodative monetary policy and fixed exchange rate regimes (Coenen et al., 2012; Christiano et al., 2011; Ilzetzki et al., 2013; Miyamoto et al., 2018).
  - Nakamura and Steinsson (2014) show fiscal multipliers at the US state level are larger than the aggregate national multiplier, reaching "about 1.5".

### Novel data, identification, and empirical approach
- Data innovation:
  - a novel, high-frequency dataset on defense procurement for EU countries constructed to analyze macro impacts and heterogeneity across procurement components.
  - this is the first study to use defense procurement spending to analyze macro impacts in all countries within the EU.
  - procurement timing (awarded contracts) is used as fiscal news to better align with private-sector expectations and isolate exogenous policy-relevant variation.
- Identification:
  - the paper employs a shift–share instrumental-variables strategy (Nakamura and Steinsson, 2014; Nunn and Qian, 2014): instrument defense procurement spending in an individual country using a combination of overall European spending and each country’s physical distance to a major adversary (distance between common military alliances and treaties between each pair of countries).
- Empirical method:
  - local projection (LP) analysis (Jordà, 2005) is used to estimate dynamic responses for horizons h = 0, 1, ..., 5.
  - model variables include normalized output (푦), normalized government defense spending (푔, as a ratio of the GDP trend), and covariates 푥 following Ramey and Zubairy (2018), Miyamoto et al. (2019), and Sheremirov and Spirovska (2022) (lags of defense spending, lags of normalized government consumption, lags of GDP, and an armed conflict index).
  - country fixed effects (훼) and time fixed effects (훿) are included; GDP trend estimated as a quadratic time polynomial.
  - direct cumulative multipliers obtained via the one-step approach of Ramey and Zubairy (2018).

### Transmission channels and heterogeneity
- Transmission channels explored include:
  - increases in consumption, investment, employment, and output documented in related studies (Ben Zeev et al., 2025).
  - a role for higher R&D spending and TFP growth in amplifying output effects.
- Heterogeneity:
  - systematic evidence provided on heterogeneity of defense spending multipliers across economic and structural environments, particularly import intensity, financing conditions, and public investment efficiency—consistent with state-dependent fiscal multiplier literature (Coenen et al., 2012; Ilzetzki et al., 2013; Abiad et al., 2016).
  - cross-border trade spillovers from defense spending are quantified, contributing to spillover literature (Auerbach and Gorodnichenko, 2013; Beetsma and Giuliodori, 2011).

### Robustness and supplementary results
- Robustness checks:
  - results robust to removing time-fixed effects (multipliers tend to increase), controlling for revenue and public debt, alternative lag structures (1, 2, and 3 lags), inclusion of country-specific time trends, leave-one-out exercises by country, and alternative subsets of EU countries (bordering non-EU countries vs. rest, and only EU member states joining from 2004).
  - comparable results obtained when restricting to pre-COVID sample as well as pre- and post-2006 samples.
- Additional comparisons:
  - baseline multipliers are broadly consistent with Sheremirov and Spirovska (2022) for government consumption shocks in international panels and with Nakamura and Steinsson (2014) for US states.
  - Ben Zeev et al. (2025) obtain a cumulative multiplier of defense news for Europe "around 2" in the first 2 years; García-Serrador et al. (2025) report a peak of "about 1.6" in the first year with a decline to zero by the fifth year in their results.
- Precision over horizons:
  - confidence intervals widen at longer horizons, making long-term estimates less precise and statistically insignificant over the long term.

*Source: IMF Working Papers — Macroeconomic Impacts of EU Defense Spending (excerpt).*

### 0.7 at the 2- and 4-year horizons documented by Ben Zeev et al. (2025) for individual European countries, and with Ilzet

### wpiea2026053-source-pdf - 0.7 at the 2- and 4-year horizons documented by Ben Zeev et al. (2025) for individual European countries, and with Ilzet

### Transmission mechanisms: demand and supply responses to defense spending
- Government consumption increases by about 0.5 percent of trend GDP in response to a defense spending shock.
- Private consumption rises by about 1 percent of trend GDP (crowd-in effect).
- Investment increases by about 1 percent of trend GDP and displays greater persistence than consumption.
- Net imports rise, reflecting higher domestic absorption (both defense-related imports and broader non-defense imports).
- Aggregating private consumption, investment, government consumption, and net exports yields a GDP response consistent with the above components.
- On the supply side:
  - Capital services expand as defense spending—particularly its investment component—raises the stock of productive assets.
  - Labor services increase, capturing higher employment and skill deployment in defense-linked sectors and indirect supply-chain effects.
  - Total factor productivity (TFP) shows a positive response, largely driven by cyclical components rather than long-run trend when decomposed using a quadratic time polynomial.
- Consumer prices:
  - The price level gradually rises and then plateaus at a higher level following a defense spending shock, suggesting aggregate demand effects outweigh supply-side improvements and producing a level shift in the price level (with increasingly imprecise estimates at longer horizons).

### Heterogeneity of the defense multiplier
- Empirical framework extends local projections by interacting cumulative defense spending with a state variable F that captures regime characteristics (e.g., import intensity, financing conditions, public investment efficiency).
- State variable definition:
  - F = 1 corresponds to favorable conditions: low import intensity, low sovereign spreads, and high public investment efficiency (low efficiency gap).
  - Threshold z* is chosen as the sample average across country–year observations.
- Import intensity:
  - Defense spending multipliers are significantly lower in periods and countries with high import intensity.
  - The output multiplier one year after a defense spending increase is around 3 in low–import-intensity environments, compared with about 1 in high–import-intensity environments.
- Fiscal space and borrowing costs:
  - Defense spending multipliers are smaller in high-spread environments (proxied by long-term sovereign bond yield spreads vis-à-vis the German long-term bond yield).
  - When sovereign spreads are low, defense spending shocks generate larger and more persistent output responses.
- Public investment efficiency:
  - Using the IMF (2025) database of investment efficiency gaps (1980–2023), defense spending multipliers are significantly larger in periods and countries with higher investment efficiency.
  - Reported estimates: a multiplier of 3 in the first three years since the shock under high investment efficiency versus a multiplier of 1 under low investment efficiency.
- Sign of the shock:
  - Identification of unexpected defense spending shocks uses residuals from estimating defense spending on controls (as in Ben Zeev, Ramey, and Zubairy (2023)), then replacing realized defense spending with the estimated shock.
  - The difference between responses to negative (unexpected cuts) versus positive (unexpected increases) shocks is neither persistent nor statistically significant, suggesting limited evidence of systematic asymmetry between positive and negative defense spending shocks.
- Robustness:
  - Results are robust to allowing country-specific slopes on the global geopolitical risk measure of Caldara and Iacoviello (2022) and to specifications interacting leads and lags of global geopolitical risk with country fixed effects.
  - Findings remain when using alternative decompositions (e.g., HP filter for TFP trend) and alternative threshold modeling (smooth transition approach).

### Cross-border spillovers: imports and output transmission via trade
- Defense spending and imports:
  - A defense spending shock of 1 percent of trend GDP leads to an immediate increase in total imports of goods and services by about 2 percent of trend GDP.
  - Decomposition (approximated using country-specific share of goods imports from EU partners):
    - Intra-EU imports increase by roughly 1 percent of trend GDP.
    - Extra-EU imports increase by roughly 1 percent of trend GDP.
  - Extra-EU imports increase both from the United States and from other non-EU countries, with a somewhat larger response from other non-EU countries.
    - About 5 percent of EU imports are from the United States, compared with 35 percent of imports from other countries (context on shares).
  - Arms imports (SIPRI unit of trend-indicator value, TIV):
    - Arms imports increase by 50 percent following a defense spending shock of 1 percent of GDP, with significant responses for both intra-EU and extra-EU sources.
- Output spillovers via trade:
  - Construction of a trade-weighted foreign defense spending shock S (weighted by intra-EU exports).
  - Domestic output response to foreign defense spending:
    - A 1 percent of GDP increase in defense spending in trading partners boosts domestic GDP, on average, by 0.2 percent on impact and 0.4 percent after 1 year.
    - This corresponds to about 15 to 25 percent of the estimated impact of domestic shocks.
  - Domestic defense spending effects remain similar when controlling for trading-partner defense spending, indicating domestic estimates are not driven by omitted foreign demand or correlated defense spending across EU countries.
  - Magnitudes align with prior findings (e.g., Beetsma and Giuliodori (2011) finding ~0.35 percent spillover from a 1 percent of GDP increase in government purchases).

### High-frequency defense procurement shocks and data construction
- Motivation for high-frequency procurement approach:
  - Monthly military procurement obligations (contract awards) provide stronger identification than annual outlays because procurement announcements constitute fiscal news that influence expectations well before payments occur.
  - Procurement obligations typically translate into payments three to four quarters after the award; relying on outlays risks downward bias due to anticipation, measurement lags, and implementation timing endogeneity.
  - Monthly awarded defense and security contracts capture unexpected policy-relevant variation and better align timing with private-sector expectations.
  - Monthly data allow estimation of impacts over short horizons (within a year) and permit sectoral disaggregation.
- Data construction and sources:
  - Collaboration with Taiyo.AI to collect procurement contract data from public sources.
  - Primary data source: Opentender, covering procurement records for 35 jurisdictions in Europe (28 EU member states, Norway, the EU Institutions, Iceland, Switzerland, Georgia, Serbia, and North-Macedonia), including TED and national procurement portals.
  - Contract-level dataset includes timing of award notice, contract value, contracting authority identity, procedural characteristics, CPV-based purpose (8-digit CPV), place of performance (NUTS or postal codes), seller information (name, postal address NUTS, nationality).
  - CPV classification: 220 8-digit CPV codes related to defense and security spending identified using European Commission’s Directive 2009/81 CPV list (reported in Table 43 of the document).
  - Institutional features: EU procurement disclosure requirements (e.g., Directive 2009/81/EC) and centralized TED reduce cross-country measurement error and reporting heterogeneity, making EU procurement-based fiscal news empirically advantageous.

*IMF Working Paper: Macroeconomic Impacts of EU Defense Spending (selected sections from the source PDF).*

### 1. Construction encompasses military and security infrastructure projects, including the development of

### wpiea2026053-source-pdf - 1. Construction encompasses military and security infrastructure projects, including the development of

### Classification framework for defense procurement
- Procurement categories defined in the dataset:
  - Construction: military and security infrastructure projects, including the development of bases, facilities, and other physical assets.
  - Equipment: procurement of weapons systems, vehicles, aircraft, naval vessels, electronics, and communication technologies essential for operational readiness.
  - Services: support activities such as training programs, simulation technologies, equipment maintenance, and hazardous material disposal.
  - Research and Development (R&D): expenditures related to military innovation, including the design and testing of new technologies and strategic systems.
- Annex Figure E.2 (described in text) covers categories from construction work and IT services to pharmaceuticals, transport equipment, cultural services, and defense/security products.
- Figure 13 panel summaries:
  - Panel A: total procurement data tracks official sources closely.
  - Panel B: 2021 procurement database strongly correlated with official European Commission figures.
  - Panel C: defense procurement as a share of GDP at annual frequency for the EU began rising by 2016 and accelerated after 2022.
  - Panel D: monthly and cross-country variation; Estonia shows concentrated big impulses, France shows more uniform distribution.
  - Panel E: construction consistently represents the largest shares; other components remain relevant at 30-40 percent of total expenses each year.
  - Panel F: composition of spending varies substantially across member states (France and Estonia examples).

### Baseline allocation and identification rationale
- Baseline approach: allocate the full value of a defense contract obligation to the award date, regardless of duration or payment schedule.
  - Motivation: capture defense news shocks and unexpected fiscal information that influence private-sector expectations and economic behavior at the time of announcement (Ramey, 2011; Ramey and Zubairy, 2018).
  - Aligns shock timing with the award date to isolate when new policy-relevant information enters the economy and when agents update expectations.

### Validity checks and sample construction
- Sample period: 2009:m1 to 2023:m12; includes EU-27 countries.
- Dependent variable construction: cumulative sum over monthly real GDP using industrial production and economic sentiment indicator in a state-space setting (Stock and Watson, 2010), normalized by trend RGDP as in Ramey and Zubairy (2018).
- Trend GDP estimation: fourth-degree polynomial for the logarithm of real GDP for 1995:m1-2023:12, excluding the COVID-19 period (2020:m3-2021:m12).
- Robust HAC standard errors are used.
- Outlier exclusion rules (applied to months):
  - Drop month if defense procurement or total procurement exceeds 5 percent of annual GDP.
  - Drop month if monthly defense procurement as a share of trend real GDP above 12.5 percent (rationale following Auerbach, Gorodnichenko, and Murphy (2019)).
  - Drop month if monthly total procurement as a share of trend real GDP exceeds 50 percent.
- Sample consistency: same set of observations used across all horizons (h = 0,...,36).

### Estimating the defense multipliers (local projections)
- Method: local projection methodology to estimate cumulative multipliers; regress cumulative output on cumulative defense procurement spending normalized by trend RGDP.
- Controls: 4 lags of defense procurement, total government procurement, and dependent variable (normalized by trend GDP).
- Key quantitative findings:
  - Cumulative fiscal multiplier remains statistically significant after 12 months.
  - Multiplier values:
    - Approximately 0.7 at 2-years (24 months).
    - About 0.8 after 3-years (36 months).
  - Exchange rate effects:
    - A 1 percent increase in defense spending leads to cumulative real depreciation of the effective exchange rate of about 0.4 percent relative to trend at 2-year horizon.
    - About 0.7 percent depreciation at 3-year horizon.
  - Confidence intervals: figures present 68% and 90% confidence intervals using robust HAC standard errors.

### Endogeneity concerns and IV (shift–share) strategy
- Endogeneity channels discussed:
  - Defense spending may be endogenous to the business cycle or political union dynamics (e.g., smaller countries timing spending when convenient; countries increasing spending when growth relaxes fiscal constraints; spending increases during heightened geopolitical risk associated with weaker economic conditions).
  - Direction of bias is not clear a priori.
- Instrumental-variable approach: shift–share (Bartik-like) instrument constructed as interaction between:
  - Shift: total EU-27 defense and security procurement normalized by EU trend GDP (captures aggregate EU-wide defense shocks).
  - Country-specific exposure shares (predetermined geographic characteristics):
    - Main exposure: geodesic population-weighted distance to Russia (Russia chosen based on ATOP-derived treaty portfolio similarity in pre-estimation period 1999-2008; Russia emerges as geopolitically farthest from EU-27 among non-EU-non-NATO members).
    - Secondary exposure: dummy equal to one if country shares a land border with any non-EU, non-NATO members (Russia, Belarus, Ukraine, Moldova, North Macedonia, Montenegro, Bosnia, Kosovo or Serbia) as of 2015; Switzerland excluded.
- Instrument construction and assumptions:
  - Country shares are based on predetermined geography and assumed unrelated to contemporaneous government policies or economic conditions.
  - Time fixed effects absorb global factors potentially correlated with EU spending; country fixed effects account for country-specific factors correlated with shares.
  - Recent literature allows Bartik-like validity requiring only one component to be exogenous (Goldsmith-Pinkham et al., 2020; Borusyak et al., 2022).
- Additional controls:
  - Global geopolitical risk index from Caldara and Iacoviello (2022), interacted with the two country-specific exposure shares to allow heterogeneous effects of worldwide geopolitical tensions across countries.

### IV estimation results and instrument diagnostics
- Visual evidence (Figure 15): countries closer to Russia and those sharing land borders with non-EU, non-NATO members exhibit, on average, higher defense procurement spending as a share of GDP.
- First-stage diagnostics (Table 3):
  - Instruments are relevant and statistically significant for all specifications.
  - Kleibergen–Paap rk Wald F statistic consistently exceeds the conventional threshold of 10 across horizons, including impact and at 24 and 36 months.
  - Hansen J overidentification test is not rejected at any horizon.
- IV estimates (Figure 16):
  - Cumulative fiscal multipliers obtained via IV are more than twice as large as those from OLS, underscoring importance of addressing endogeneity.
  - Solid lines depict point estimates; dotted lines represent 68% and 90% confidence intervals based on robust HAC standard errors.

*Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026053-source-pdf.pdf*

### 1.5 at the 12-month horizon, approximately 1.7 at two years, and around 1.9 after three years, both statistically

### 1.5 at the 12-month horizon, approximately 1.7 at two years, and around 1.9 after three years, both statistically

### Key findings
- Estimated defense-spending multipliers:
  - 1.5 at the 12-month horizon
  - approximately 1.7 at two years
  - around 1.9 after three years
- A 1 percent increase in defense spending leads to a cumulative real depreciation of the effective exchange rate of roughly 4 percent relative to trend at the two- and three-year horizons.
- OLS estimates yield much smaller multipliers, suggesting a downward bias from reverse causality and omitted-variable concerns.
- Aggregate IV estimate: roughly 1.9 at 36 months.
- Heterogeneity by country characteristics: multipliers tend to be larger in environments with low import intensity, available fiscal space, and high public investment efficiency.

### Robustness checks
- Varying the number of lags in the local projection framework to test dynamic response sensitivity.
- Re-estimating models with and without time fixed effects to control for global movements in defense spending; coefficients remain stable across these specifications.
- Leave-one-out analysis by sequentially removing each country from the sample to check for influential observations.
- Restricting the sample to the 2009–2019 period (excluding the COVID-19 recession and the geopolitical shock associated with Russia’s invasion of Ukraine in 2022); results persist in this window.
- Annex Figure H.1 (referenced) presents detailed robustness results.
- Robust inference to weak instruments (conditional likelihood ratio (CLR) test proposed by Finlay and Magnusson (2009)) increases confidence intervals but remains in close range with baseline CIs, supporting identification.

### Heterogeneous impacts by spending category
- Procurement decomposed into four sub-components using CPV codes: construction, services, equipment, and R&D. Four mutually exclusive dummies identify which sub-component is dominant in each month.
- Changes in composition between 2009 and 2023:
  - Construction averaged nearly 70% of procurement overall, declining from a peak of 80% in 2011 to around 55% by 2023.
  - R&D increased to about 20% by 2023.
  - Services rose to 12–15% by 2023.
  - Equipment remained the smallest component throughout but increased recently.
- OLS cumulative GDP responses to a 1 percent of GDP increase in defense procurement by component:
  - Equipment-biased packages: largest relative medium-term multipliers, reaching about 2.7 at 36 months.
  - Services packages: multipliers around 1 after 24 months and 1.2 after three years.
  - Construction-biased packages: multipliers near 0.6 at 36 months; strongly statistically significant.
  - R&D-related packages: cumulative multiplier near 1.0 at the 36-month horizon; estimates statistically significant at the 68% level.
- IV implementation for subcomponents:
  - Results similar to OLS; equipment spending stands out as the dominant driver of output effects relative to the aggregate multiplier.
  - For subcomponents, each of the four mutually exclusive dummies is interacted with the two shift–share instruments, creating eight instruments instead of two; the endogenous variable becomes the cumulative change in defense procurement interacted with the relevant dummy.

### Methodology notes
- Baseline specification uses cumulative defense procurement and cumulative real GDP, both normalized by trend GDP, with mutually exclusive dummies for the dominant subcomponent in each month.
- Cumulative variables defined as:
  - g_{i,t+h,t0} = sum_{s=t0}^{t+h} g_{i,t+s}
  - y_{i,t+h,t0} = sum_{s=t0}^{t+h} y_{i,t+s}
- Identification relies on IV shift–share instruments and local projections; CLR test used for robustness to weak instruments.

### Conclusion and policy implications
- Increased defense outlays generate positive effects on output and induce exchange-rate depreciation effects; intra-EU spillovers through trade are positive and economically meaningful.
- Procurement composition and timing materially shape macroeconomic outcomes; equipment-intensive packages yield the largest multipliers.
- Structural characteristics (import intensity, fiscal space, public investment efficiency) significantly affect multiplier size and fiscal outcomes (e.g., public debt dynamics when fiscal space is limited).
- Caveat: the ongoing coordinated and large-scale increase in defense spending across Europe differs from the 1989–2023 sample; multipliers relevant to the current coordinated buildup might be lower than historical estimates if monetary policy is less accommodative.
- Suggested avenues for future research include:
  - Alternative classifications of defense spending (e.g., automated contract parsing using large language models, keyword-based techniques).
  - Allocating expenditures uniformly over contract durations rather than assigning full impact at award date to distinguish anticipated from unanticipated shocks.
  - Incorporating supplier-level information and buyer-location/place-of-performance data to map geographic and sectoral spillovers.
  - Jointly identifying shocks to defense and non-defense spending to compare respective multipliers.

*IMF WORKING PAPERS Macroeconomic Impacts of EU Defense Spending — INTERNATIONAL MONETARY FUND*

### References

### References

### Bibliography
- Abiad, A., Furceri, D., & Topalova, P. (2016). The macroeconomic effects of public investment: Evidence from advanced economies. Journal of Macroeconomics, 50, 22ce4-240.
- Acalin, M. J., Alonso-Albarran, V., Arroyo, C., Lam, R., Martinez, M., Nguyen, A. D., Roch, F., Sher, G., & Solovyeva, A. (2025). Fiscal Guardrails Against High Debt and Looming Spending Pressures. International Monetary Fund Staff Discussion Notes 2025/004.
- Alesina, A., Favero, C., & Giavazzi, F. (2015). The output effect of fiscal consolidation plans. Journal of International Economics, 96, S19-S42.
- Alliance Treaty Obligations and Provisions (ATOP) Project. (2023). http://www.atopdata.org/
- Alloza, M., Burriel, P., & Pérez, J. J. (forthcoming). The macroeconomic effects of defense procurement contracts: Evidence from Spain. Banco de España Working Paper.
- Antolin-Diaz, J., & Surico, P. (2025). The long-run effects of government spending. American Economic Review, 115(7), 2376-2413.
- Antonova, A., Luetticke, R., & Muller, G. J. (2025). The Military Multiplier. CESifo Working Paper (No. 11882).
- Auerbach, A. J., & Gorodnichenko, Y. (2012). Fiscal multipliers in recession and expansion. In A. Alesina & F. Giavazzi (Eds.), Fiscal Policy after the Financial Crisis (pp. 63–98). University of Chicago Press.
- Auerbach, A. J., & Gorodnichenko, Y. (2013). Output spillovers from fiscal policy. American Economic Review, 103(3), 141–146.
- Auerbach, A. J., Gorodnichenko, Y., & Murphy, D. (2019). Local fiscal multipliers and fiscal spillovers in the United States. IMF Economic Review, 67(1), 7–39.
- Barro, R. J., & Redlick, C. J. (2011). Macroeconomic effects from government purchases and taxes. The Quarterly Journal of Economics, 126(1), 51–102.
- Beetsma, R., & Giuliodori, M. (2011). The effects of government purchases shocks: Review and estimates for the EU. Economic Journal, 121(550), F4–F32.
- Ben Zeev, N., Ramey, V. A., & Zubairy, S. (2023). Do government spending multipliers depend on the sign of the shock?. In AEA Papers and Proceedings (Vol. 113, pp. 382-387).
- Ben Zeev, N., Pappa, E., & Scola Gagliardi, E. (2025). The offensive power of defense news in Europe. CEPR Discussion Paper No. 20637.
- Benmelech, E., & Monteiro, J. (2025). Military Spending and War. National Bureau of Economic Research (No. w34123).
- Bokan, Nikola; Pascal Jacquinot; Magdalena Lalik; Georg Müller; Romanos Priftis; and Rodolfo Rigato (2025). “Macroeconomic impacts of higher defence spending: a model-based assessment” ECB Economic Bulletin, Issue 6/2025.
- Borusyak, K., Hull, P., & Jaravel, X. (2022). Quasi-experimental shift-share research designs. Review of Economic Studies, 89(1), 181–213.
- Briganti, E., Brunet, G., & Sellemi, V. (2025). When Does Government Spending Matter? It’s All in the Measurement. Working Paper.
- Cacciatore, M., R. Duval, D. Furceri & A. Zdzienicka (2021). Fiscal multipliers and job-protection regulation, European Economic Review, volume 132, 103616.
- Caldara, D., & Iacoviello, M. (2022). Measuring geopolitical risk. American Economic Review, 112(4), 1194–1225.
- Canova, F., & Pappa, E. (2025). The macroeconomic effects of EU regional Structural Funds. Journal of the European Economic Association, 23(1), 327-360.
- Chiba, D., Johnson, J. C., & Leeds, B. A. (2015). Careful commitments: Democratic states and alliance design. Journal of Politics, 77(4), 968–982.
- Choi, S., Willems, T., & Yoo, S. Y. (2024). Revisiting the monetary transmission mechanism through an industry-level differential approach. Journal of Monetary Economics, 145, 103556.
- Christiano, L. J., Eichenbaum, M., & Rebelo, S. (2011). When is the government spending multiplier large? Journal of Political Economy, 119(1), 78–121.
- Coenen, G., Erceg, C. J., Freedman, C., Furceri, D., Kumhof, M., Lalonde, R., Laxton, D., Lindé, J., Mourougane, A., Muir, D., Mursula, S., de Resende, C., Roberts, J., Roeger, W., Snudden, S., Trabandt, M., & in 't Veld, J. (2012). Effects of fiscal stimulus in structural models. American Economic Journal: Macroeconomics, 4(1), 22–68.
- Collier, P. (2006). War and military spending in developing countries and their consequences for development. The Economics of Peace and Security Journal, 1(1).
- Corsetti, G., Meier, A., & Müller, G. J. (2012). What determines government spending multipliers? Economic Policy, 27(72), 521–565.
- Cox, L., Müller, G. J., Pasten, E., Schoenle, R., & Weber, M. (2024). Big g. Journal of Political Economy, 132(10), 3260-3297.
- European Investment Bank (EIB). (2025). Innovation integration and simplification in Europe. Chapter 2. EIB Investment Report 2024/2025.
- Finlay, K. and Magnusson, L.M., 2009. Implementing weak-instrument robust tests for a general class of instrumental-variables models. The Stata Journal, 9(3), pp.398-421.
- Garcia-Serrador, A., Sarasa-Flores, D., & Ulloa, C. (2025). Buy guns or buy roses?: EU defence spending fiscal multipliers. BBVA Research Working Paper 25-06.
- Gleditsch, N. P., Wallensteen, P., Eriksson, M., Sollenberg, M., & Strand, H. (2002). Armed conflict 1946-2001: A new dataset. Journal of peace research, 39(5), 615-637.
- Goldsmith-Pinkham, P., Sorkin, I., & Swift, H. (2020). Bartik instruments: What, when, why, and how. American Economic Review, 110(8), 2586–2624.
- Hall, R. E. (2009). By how much does GDP rise if the government buys more output? Brookings Papers on Economic Activity, 40(2), 183–231.
- Hooker, M. A., & Knetter, M. M. (1997). The effects of military spending on economic activity: Evidence from state procurement spending. Journal of Money, Credit and Banking, 29(3), 400–421.
- Ilzetzki, E. (2025). Guns and growth: The economic consequences of defense spending buildups. Kiel Report No. 2.
- Ilzetzki, E., Mendoza, E. G., & Végh, C. A. (2013). How big (small?) are fiscal multipliers? Journal of Monetary Economics, 60(2), 239–254.
- International Monetary Fund (IMF). (2025). Fiscal Monitor: Spending Smarter—How Efficient and Well-Allocated Public Spending Can Boost Economic Growth. Washington, DC.
- Jordà, Òscar. 2005. Estimation and inference of impulse responses by Local Projections. American Economic Review 95(1): 161–182.
- Miyamoto, W., Nguyen, T. L., & Sergeyev, D. (2018). Government spending multipliers under the zero lower bound: Evidence from Japan. American Economic Journal: Macroeconomics, 10(3), 247-277.
- Miyamoto, W., Nguyen, T. L., & Sheremirov, V. (2019). The effects of government spending on real exchange rates: Evidence from military spending panel data. Journal of International Economics, 116, 144-157.
- Nakamura, E., & Steinsson, J. (2014). Fiscal stimulus in a monetary union: Evidence from US regions. American Economic Review, 104(3), 753–792.
- Nekarda, C. J., & Ramey, V. A. (2011). Industry evidence on the effects of government spending. American Economic Journal: Macroeconomics, 3(1), 36–59.
- Nunn, N., & Qian, N. (2014). US food aid and civil conflict. American Economic Review, 104(6), 1630–1666.
- Opentender. (2023). European public procurement data. https://opentender.eu/
- Ramey, V. A. (2011). Identifying government spending shocks: It's all in the timing. Quarterly Journal of Economics, 126(1), 1–50.
- Ramey, V. A. (2016). Macroeconomic shocks and their propagation. Handbook of Macroeconomics, 2, 71–162.
- Ramey, V. A., & Shapiro, M. D. (1998). Costly capital reallocation and the effects of government spending. Carnegie-Rochester Conference Series on Public Policy, 48, 145–194.
- Ramey, V. A., & Zubairy, S. (2018). Government spending multipliers in good times and in bad: Evidence from US historical data. Journal of Political Economy, 126(2), 850–901.
- Romer, Christina D., and David H. Romer. 2017. New Evidence on the Aftermath of Financial Crises in Advanced Countries. American Economic Review 107 (10): 3072–3118.
- Sheremirov, V., & Spirovska, S. (2022). Fiscal multipliers in advanced and developing countries: Evidence from military spending. Journal of Public Economics, 208, 104631.
- Signorino, C. S., & Ritter, J. M. (1999). Tau-b or not tau-b: Measuring the similarity of foreign policy positions. International Studies Quarterly, 43(1), 115–144.
- Stock, J. H., & Watson, M. W. (2010). Monthly GDP and GNI-research memorandum. Manuscript, Princeton University.
- Stockholm International Peace Research Institute (SIPRI). (2023). SIPRI Military Expenditure Database. https://www.sipri.org/databases/milex
- Tenders Electronic Daily (TED). (2023). Supplement to the Official Journal of the EU. https://ted.europa.eu/

### Figures

### Figure captions and notes
- Figure 1. Military Spending in EU 1973-2023 (Percent of GDP)
  - Source: SIPRI.
- Figure 2. Response to a 1 Percent of Trend GDP Increase in Defense Spending
  - Panels: A. Defense Spending (Percent of trend GDP); B. Output (Percent of trend GDP); C. Cumulative Multiplier.
  - Note: Responses to a positive defense spending shock of 1 percent of GDP in a panel of 27 EU countries. Panel C is the cumulative multiplier estimated based on Equation (2). Standard errors are clustered by country. Dashed blue lines indicate 68% confidence intervals; dashed black lines indicate 90% confidence intervals.
- Figure 3. Response of Short-Term Interest Rate and Real Effective Exchange Rate to a 1 Percent of Trend GDP Increase in Defense Spending
  - Panels: A. Short-Term Interest Rate (Percentage Points); B. Real Effective Exchange Rate (Percent).
  - Note: Extends baseline specification in Equation (1), controls for own lags and baseline controls. Blue lines indicate 68% confidence intervals; black lines indicate 90% confidence intervals.
- Figure 4. Responses of Demand Components to a 1 Percent of Trend GDP Increase in Defense Spending (Percent of trend GDP)
  - Panels: A. Government Consumption; B. Private Consumption; C. Investment; D. Imports; E. Exports.
  - Note: Responses to a positive defense spending shock of 1 percent of GDP in a panel of 27 EU countries. Standard errors clustered by country. Blue lines indicate 68% confidence intervals; black lines indicate 90% confidence intervals.
- Figure 5. Responses of Supply Components (Percent)
  - Panels: A. Capital Service; B. Labor Service Supply; C. Total Factor Productivity.
  - Note: Responses to a positive defense spending shock of 1 percent of GDP in a panel of 27 EU countries. Standard errors clustered by country. Blue lines indicate 68% confidence intervals; black lines indicate 90% confidence intervals.
- Figure 6. Response of TFP Components (Percent)
  - Panels: A. TFP Cyclical Component; B. TFP Trend Component.
  - Note: Responses to a positive defense spending shock of 1 percent of GDP in a panel of 27 EU countries. Standard errors clustered by country. TFP trend is estimated from estimated from a quadratic polynomial trend. Blue lines indicate 68% confidence intervals; black lines indicate 90% confidence intervals.
- Figure 7. Response of Price (Percent)
  - Note: Responses to a positive defense spending shock of 1 percent of GDP in a panel of 27 EU countries. Standard errors clustered by country. Inflation winsorized to limit influence of outliers; a new price index is built from winsorized series. Regressed log of price index following Equation (1), controlling for own lags and baseline controls. Blue lines indicate 68% confidence intervals; black lines indicate 90% confidence intervals.
- Figure 8. Heterogeneity of Defense Multiplier (Threshold approach)
  - Panels: A. Import intensity — Output cumulative multipliers Differences: Low vs. High Import Intensity; B. Financial Conditions — Output cumulative multipliers Differences: Low vs. High Spread; C. Public investment efficiency — Output cumulative multipliers Differences: High vs. Low Efficiency.
  - Notes: Left chart shows cumulative multiplier under two environments; right chart shows differences between “blue” and “red” lines. Blue lines indicate 68% confidence intervals and black lines 90% confidence intervals. Shaded areas describe 68/90 percent confidence intervals for corresponding state.
- Figure 9. Differences Between Cumulative Multipliers of Negative Shocks and Positive Shocks
  - Note: Figure shows differences between the cumulative multipliers of negative shocks versus that of positive shocks. Bounds are 68 percent and 90 percent confidence intervals.
- Figure 10. Responses of Imports (Percent of trend GDP)
  - Panels: A. Total imports; B. Total Imports from Intra-EU; C. Total Imports from Extra-EU; D. Total imports from non-US Extra EU; E. Total Imports from US.
  - Note: Responses to a positive defense spending shock of 1 percent of GDP in a panel of 27 EU countries. Standard errors clustered by country. Blue lines indicate 68% confidence intervals; black lines indicate 90% confidence intervals.
- Figure 11. Impulse Responses of Arms Imports to a Defense Spending Shock (Percent)
  - Panels: A. Arms Imports: Total; B. Arms Imports: Intra EU; C. Arms Imports: Extra EU.
  - Note: Responses to a positive defense spending shock of 1 percent of GDP in a panel of 27 EU countries. Standard errors clustered by country. Blue lines indicate 68% confidence intervals; black lines indicate 90% confidence intervals.
- Figure 12. Responses of GDP to Defense Spending Shock in Trading Partner Versus Domestic Shock (Percent of trend GDP)
  - Panels: A. Spillover defense spending shock; B. Domestic defense spending shock.
  - Note: Left chart: Responses to a positive military spending shock of 1 percent of trading partner GDP in a panel of 27 EU countries. Response rescaled by average export share. Right chart: Response to domestic defense spending shock (1 percent of domestic GDP).
- Figure 13. OpenTender Procurement Database vs Official Aggregates
  - Panels: A. EU’s Total Procurement (% of GDP); B. Comparison: Our sample vs EC’s (2019 – 2021); C. EU’s Defense Procurement (% of GDP); D. Monthly’s Defense Procurement: Selected Countries; E. Defense and Security Procurement by Subcomponents; F. Defense and Security Procurement by Subcomponents: Selected countries.
  - Source: Taiyo and European Commission and Staff estimate.
  - Note: To define defense and security procurement we follow EU Common Procurement Vocabulary, adopted by Regulation EC No. 213/2008 and Taiyo. The US averages are based on Cox et al. (2024).
- Figure 14. Defense and Security Procurement: OLS Results
  - Panels: A. GDP; B. REER.
  - Note: Baseline OLS estimates from Equation (8), dynamic cumulative response to a defense and security procurement shock of 1 percent of trend GDP for EU-27 countries. Panel A: cumulative response of normalized GDP. Panel B: cumulative normalized response of the REER (real effective exchange rate, where a negative value indicates depreciation). Standard errors robust (HAC). Blue lines indicate 68% confidence intervals; black lines indicate 90% confidence intervals.
- Figure 15. Relevance of Instruments
  - Panels: A. IV1: Distance to Russia; B. IV2: Non-EU-non-NATO member Border.
  - Note: Demonstrates relevance of instruments defined in Equation (8). Panel A: population-weighted average distance to Russia (km) and average defense and security procurement as share of trend GDP. Panel B: average defense and security procurement as share of trend GDP, grouped by whether country borders a non-EU, non-NATO member (indicator = 1 if yes).
- Figure 16. Defense and Security Procurement: IV Results
  - Panels: A. GDP; B. REER.
  - Note: Baseline IV estimates from Equation (9) and (10) using instruments defined in Equation (8), dynamic cumulative response to a defense and security procurement shock of 1 percent of trend GDP for EU-27 countries. Standard errors robust HAC (4). Blue lines indicate 68% confidence intervals; black lines indicate 90% confidence intervals.
- Figure 17. Evolution Over Time of Subcomponents of Defense and Security Procurement Packages
  - Note: Chart shows, for each month in sample, the average share of EU-27 countries where a specific procurement subcomponent (Equipment, Services, Construction, or R&D) is dominant. Construction consistently has the highest average dominance across countries, while Equipment, Services, and R&D remain less prevalent.
- Figure 18. Heterogeneity by Defense and Security Procurement Subcomponents: OLS Results
  - Panels: A. Equipment; B. Services; C. Construction; D. R&D.
  - Note: OLS estimates from Equation (11), dynamic cumulative response of GDP to a subcomponent defense and security procurement shock of 1 percent of trend GDP for EU-27 countries. Standard errors robust HAC (4). Blue lines indicate 68% confidence intervals; black lines indicate 90% confidence intervals.
- Figure 19. Heterogeneity by Defense and Security Procurement Subcomponents: IV Results
  - Panels: A. Equipment; B. Services; C. Construction; D. R&D.
  - Note: IV estimates from Equation (11), dynamic cumulative response of GDP to subcomponent procurement shocks. For subcomponents, each category dummy is interacted with two shift–share instruments from Equation (3), creating eight instruments. Standard errors robust HAC (4). Blue lines indicate 68% confidence intervals; black lines indicate 90% confidence intervals.

### Tables

### Table 1. Exogeneity Test of Defense Spending
- Table presents estimates from regression of military spending. All specifications control for country and time fixed effects. Standard errors clustered by country are in parentheses.
- Reported sample coefficients (selection shown in table):
  - Output (-1): 0.005 (0.018) ; 0.017 (0.012)
  - Output (-2): 0.003 (0.008) ; -0.006 (0.006)
  - Government consumption (-1): -0.045 (0.052) ; -0.067 (0.058)
  - Government consumption (-2): 0.014 (0.034) ; 0.043 (0.035)
  - Armed conflict index: 0.036* (0.018) ; 0.056*** (0.001)
  - PCA: -0.000 (0.000)
  - Geopolitical Risk: 0.001 (0.001)
  - Monetary Policy Shocks: -0.000 (0.000)
- Note: Geopolitical risk available for 12 EU countries from Caldara and Iacoviello (2022). PCA: principal component of output growth, inflation, and gov expenditure as percent of GDP at year t, projected in October previous year from the WEO vintage database. Monetary policy shock is from Choi et al. (2024). ***p<0.01, **p<0.05, *p<0.1

### Table 2. Country Contribution to Total EU Defense and Security
- Note: Shows each EU-27 country’s contribution to total EU-27 Defense and security procurement spending.
- Country statistics (Mean, P50, P25, P75, SD):
  - Austria: 2.4 1.5 0.7 2.9 3.9
  - Belgium: 4.8 3.5 2.3 5.5 4.1
  - Bulgaria: 2.3 1.5 1.0 2.4 3.2
  - Croatia: 0.5 0.3 0.1 0.6 0.6
  - Cyprus: 0.1 0.0 0.0 0.2 0.3
  - Czech Republic: 5.0 4.0 2.6 5.6 3.9
  - Denmark: 6.9 6.1 3.8 8.8 4.1
  - Estonia: 0.4 0.1 0.0 0.3 0.8
  - Finland: 6.3 5.5 3.5 8.1 4.9
  - France: 19.6 17.1 14.1 24.1 9.0
  - Germany: 6.4 5.2 3.4 8.2 4.6
  - Greece: 0.6 0.3 0.1 0.7 0.9
  - Hungary: 2.5 1.7 1.1 3.1 2.4
  - Ireland: 2.0 0.8 0.3 2.1 3.6
  - Italy: 5.2 3.3 1.6 7.2 5.8
  - Latvia: 0.9 0.6 0.2 1.0 1.2
  - Lithuania: 0.7 0.2 0.1 0.7 1.2
  - Luxembourg: 0.5 0.2 0.0 0.5 0.9
  - Malta: 0.0 0.0 0.0 0.0 0.2
  - Netherlands: 4.7 3.0 1.6 5.6 5.4
  - Poland: 13.6 11.5 7.6 16.0 9.6
  - Portugal: 0.6 0.3 0.1 0.6 1.2
  - Romania: 2.7 1.9 1.1 3.1 3.0
  - Slovakia: 1.8 1.1 0.7 2.3 1.7
  - Slovenia: 0.8 0.6 0.2 1.1 1.1
  - Spain: 5.9 5.2 3.3 7.0 4.5
  - Sweden: 3.0 2.2 1.0 3.9 2.9

### Table 3. First Stage Statistics
- Note: First-stage statistics for instrument strength (Cragg-Donald and Kleibergen-Paap rk Wald statistics) and the Hansen J p-value for overidentification (testing instrument validity) for Equation (10).
- Horizon, Cragg-Donald Wald F-statistic, Kleibergen-Paap rk Wald F, Over-identification Hansen J p-value:
  - h = 0: 28.28 10.68 0.90
  - h = 1: 16.93 10.16 0.64
  - h = 2: 13.85 7.58 0.45
  - h = 3: 17.96 7.15 0.38
  - h = 4: 20.28 7.20 0.48
  - h = 5: 18.83 6.75 0.51
  - h = 6: 16.39 5.71 0.53
  - h = 7: 14.90 5.06 0.55
  - h = 8: 15.77 5.42 0.58
  - h = 9: 18.10 6.37 0.55
  - h = 10: 21.46 7.39 0.53
  - h = 11: 25.45 8.50 0.52
  - h = 12: 28.71 9.00 0.55
  - h = 13: 30.81 9.47 0.51
  - h = 14: 32.83 9.99 0.57
  - h = 15: 36.70 11.07 0.48
  - h = 16: 39.58 10.82 0.49
  - h = 17: 41.18 10.44 0.45
  - h = 18: 42.40 10.68 0.39
  - h = 19: 41.78 11.01 0.38
  - h = 20: 42.07 11.08 0.36
  - h = 21: 43.57 11.84 0.37
  - h = 22: 42.64 11.69 0.35
  - h = 23: 44.68 12.65 0.34
  - h = 24: 43.90 12.38 0.34
  - h = 25: 43.41 12.27 0.34
  - h = 26: 41.39 11.75 0.36
  - h = 27: 41.22 11.79 0.36
  - h = 28: 42.27 12.31 0.35
  - h = 29: 41.54 12.21 0.34
  - h = 30: 42.80 13.25 0.30
  - h = 31: 40.22 12.81 0.28
  - h = 32: 39.11 12.63 0.26
  - h = 33: 39.52 13.00 0.23
  - h = 34: 37.91 12.76 0.20
  - h = 35: 38.52 13.44 0.17
  - h = 36: 38.73 13.72 0.16

*Italic source: wpiea2026053-source-pdf - References (IMF Working Papers Macroeconomic Impacts of EU Defense Spending).*

### Annex A. Data Sources

### Annex A. Data Sources

### Part 1: Analysis with Annual Data
- Real GDP — United Nation’s National Accounts Aggregates Database
- Real Government Consumption — United Nation’s National Accounts Aggregates Database
- Military Spending (as percent of GDP) — Stockholm International Peace Research Institute (SIPRI)
- Arms Conflict Index — Miyamoto et al. (2019), but extended using data from UCDP/PRIO
- Import — United Nation’s National Accounts Aggregates Database
- Real Private Consumption — United Nation’s National Accounts Aggregates Database
- Real Investment — United Nation’s National Accounts Aggregates Database
- Capital Service — Penn World Table (version 10.01)
- Labor Service Supply — Penn World Table (version 10.01)
- Total Factor Productivity — Penn World Table (version 10.01)
- Long term sovereign bond yield — Haver
- Short term interest rate — OECD, Haver
- Investment efficiency gap — IMF (2025)
- Bilateral trade — IMF’s Direction of Trade
- Revenue as percent of GDP — IMF WEO database
- Debt as percent of GDP — IMF WEO database
- Geopolitical Risk — Caldara and Iacovello (2022)
- Monetary policy shock — Choi et al. (2024)

### Part 2: Analysis with High-Frequency Data
- Defense and Security procurement — Opentender
- Total Procurement — Opentender
- Industrial production (excluding Construction) — Haver
- Industrial production - Manufacturing — Haver
- Economic sentiment indicator — Haver
- REER — Bank for International Settlements (BIS)
- Geopolitical risk index — Caldara and Iacovello (2022)
- Distance to Russia (pop-wt, km) — GeoDistance Database (CEPII)
- Shared borders with non-EU, non-NATO members — GeoDistance Database (CEPII)
- Geopolitical alliance — Alliance Treaty Obligations and Provisions (ATOP)
- Quarterly GDP — Haver
- Quarterly GDP deflator — Haver
- HICP — Haver

*Annex A. Data Sources — IMF Working Paper Macroeconomic Impacts of EU Defense Spending*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026053-source-pdf.pdf_