## 1. Introduction (wpiea2024098-print-pdf)

## Source details

**Canonical URL:** [1. Introduction (wpiea2024098-print-pdf)](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024098-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024098-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024098-print-pdf.pdf.json)

---

### Background and motivation
- About one quarter of the world’s population lives in conflict-affected areas; number of conflicts reaching historical highs since World War II (UN).
- Violent intrastate conflicts impose dire humanitarian costs and are associated with negative, enduring effects on economic performance; bi-directional associations create “conflict traps.”

### Hypothesis and identification
- Hypothesis: Intrastate conflicts disproportionately reduce industry growth in industries that rely more on R&D (higher technological intensity).
- Identification: Within-country variation across 2-digit manufacturing industries (ISIC Rev. 3, 15-36) based on technological intensity in the spirit of Rajan and Zingales (1998).
- Benchmark for technological intensity: industry-level R&D intensity measured using R&D expenditures of large, listed firms from the US (baseline) with robustness checks using OECD, patents, and quality certificates.

### Data scope (summary)
- Industry coverage: 2-digit manufacturing industries (ISIC 15-36).
- Country/sample coverage: 114 countries (21 advanced economies).
- Time period: 1980-2018.
- Conflict data: Major Episodes of Political Violence (MEPV) database (intrastate conflicts).
- Outcome variable: industry labor productivity (value added per employee); growth measured as log differences × 100.

### Mechanisms: why R&D-intensive industries are more affected
- R&D investment characteristics increasing vulnerability:
  - Longer delay to produce outcomes; higher risk.
  - Larger scale, lumpy initial investment; firms defer under uncertainty.
  - Highly specialized, low-redeployability, high-irreversibility assets.
- Institutional and market impairments during conflicts:
  - Weaker enforcement and intellectual property protection → appropriability problems → underinvestment in R&D.
  - Declines in consumer and investor confidence reduce demand for high-tech products.
  - Disruptions to infrastructure, collaboration, and knowledge spillovers.

### Organization
- Section 2: Data.
- Section 3: Empirical methodology.
- Section 4: Results.
- Section 5: Conclusion.

*From "1. Introduction", IMF WORKING PAPERS, INTERNATIONAL MONETARY FUND.*

---

### 2. Data

### Episodes of conflicts (MEPV)
- Event definition: systematic and sustained use of violence by organized groups with human toll > 500 deaths.
- Focus: intrastate episodes (civil and ethnic violence, wars); temporal coverage up to 2018.
- Coding: country-year dummy = 1 for years with intrastate conflicts; alternative coding assigns 1 (2) when score is below (above) the top quartile.
- Distributional facts:
  - Total events in sample: 519 events.
  - Conflicts cover about 19 percent of country-industry-year observations.
  - Geographic spread (country-year observations with conflicts): Africa 70, Asia 279, Americas 91, Europe 79.
  - 97 percent of events occurred in emerging markets and developing economies.

### Industry-level data (UNIDO)
- Labor productivity = industry value added (US$) / number of employees.
- Growth = log difference of labor productivity × 100; restricted to 100 and -100 percent (roughly 1st and 99th percentiles); winsorized checks similar.
- Control: industry value added share in manufacturing (percent).

### Industry technological intensity (main proxies)
- Baseline proxy (RDI baseline): R&D expenditures as share of overall capital expenditures; source Igan et al. (2022), Compustat US listed firms; median across firms within industry averaged 1980–1999.
  - Sample statistics: mean = 0.97; standard deviation = 1.34.
  - 25th percentile industry: paper products (ISIC 21) — RDI baseline = 0.266.
  - 75th percentile industry: machinery (ISIC 29) — RDI baseline = 0.926.
- Robustness proxies:
  - RDI alternative I: R&D expenditures to sales (Brown et al. 2017), median across firms, 1990s.
  - RDI alternative II: industry R&D expenditure share in output (OECD), averaged 1990-2009.
  - Patent intensity (NBER Patent Database, USPTO applications 1980-1986).
  - QC intensity: share of firms holding quality certificates (World Bank Enterprise Surveys, 2006–2011).
- Concordance: proxies positively correlated; industries above median/quartiles mostly overlap.
- Galindo-Rueda and Verger classification mapped to ISIC for high/medium/low-tech.

### Other industry characteristics
- Firm-level-derived measures (median across firms, 1980-1999): dependence on external finance; liquidity needs; asset tangibility.
- Industry-level measures: physical capital intensity (NBER-CES), product complexity (inverse Herfindahl of intermediate inputs), import/export exposure (OECD STAN BTDIxE).
- Selected sample values presented in Tables 1 and 3 (see source for full table entries).

### Country-level data for robustness
- Macroeconomic/financial: real GDP per capita (constant US$), credit to private sector (% of GDP), trade (% of GDP) — World Bank WDI.
- Financial openness: Chinn and Ito (2006) index (0–1); alternate FDI stock (% of GDP) from Milesi-Ferretti (2022).
- Institutional quality: constraints on executives index from Polity V (0–7).
- Real exchange rate appreciation: nominal exchange rates and CPI/PPI inflation (Ha et al. 2023).

### Sample details
- Period: main sample 1980-2018; countries: 114; industries: ISIC 15-36.
- Exclude country-industry pairs with < 5 years of data (no effect on results).
- Conflicts: 519 events; conflicts ≈ 19 percent of observations.

*Source: IMF Working Paper — Section 2. Data (wpiea2024098-print-pdf).*

---

### 3. Methodology

### Conceptual approach
- Difference-in-differences framework exploiting country-level conflict shocks interacting with industry technological intensity (Tech intensity_j × Conflict_c,t−r) for r = −2, −1, 0, 1, 2.
- Dependent variable: Δlog(LP_c,j,t) — growth rate of industry labor productivity (percentage points).
- Controls: lagged industry VA share; fixed effects: country-year, country-industry, industry-year.
- Standard errors: clustered at country-industry level; robustness to alternative clustering reported.

### Baseline econometric specification (equation (1))
- Δlog(LP_c,j,t) = α VAshare_c,j,t−1 + Σ_{r=−2}^{2} β_r TechIntensity_j × Conflict_c,t−r + θ_c,t + θ_c,j + θ_j,t + ε_c,j,t
- Interpretation:
  - β_{−2}, β_{−1}: pre-conflict dynamics (test for pre-trends).
  - β_0: contemporaneous effect in conflict years.
  - β_1, β_2: aftermath dynamics (persistence or rebound).

### Extensions
- Equation (2): adds Σ μ_r I_j × Conflict_c,t−r to control for other industry characteristics I_j (physical capital intensity, input concentration, dependence on external finance, liquidity needs, asset tangibility, import/export exposure).
- Equation (3): adds Σ γ_r TechIntensity_j × X_c,t−r to allow heterogeneous effects with country-level factors X_c,t−r (economic/financial development, openness, FDI, institutional quality).

### Identification assumptions
- Industries inherently and persistently differ in R&D reliance; US-based industry technological intensity serves as exogenous cross-country proxy.
- Pre-trend tests: β_{−2}, β_{−1} expected statistically insignificant; significant negative β_0 indicates disproportionate adverse effect on high-tech industries.

---

### 4. Results and Robustness

### Core empirical findings
- Conflicts lead to declines in labor productivity growth, disproportionately so in industries with higher technological intensity.
- Quantitative example (paper products vs. machinery):
  - RDI baseline: paper products (ISIC 21) = 0.266; machinery (ISIC 29) = 0.926.
  - Estimated additional decline in machinery relative to paper products in conflict years: 1.5 percentage points (2.2 times the difference in R&D intensity measures for those two industries).
- Pre-conflict trends: high-tech industries do not show different growth patterns pre-conflict (supporting identification).
- Post-conflict: two-year window following conflicts shows no rebound that offsets losses in high-tech industries; lost productivity not recovered in the short-term aftermath.

### Baseline estimates (selected)
- Table 4 (Tech intensity × Conflict c,t):
  - Column (1): -1.325** (0.543)
  - Column (2)/(3): -2.236*** (0.770)
- Lag/lead interaction coefficients (selected): Tech intensity × Conflict c,t+2 ≈ 0.705 (0.585); Tech intensity × Conflict c,t+1 varied; Tech intensity × Conflict c,t−1 ≈ 0.728 (0.916).
- Industry initial VA share: VA share_c,j,t−1 = -1.498*** (0.087).
- Observations = 49,518; R-squared = 0.358.
- Standard errors clustered at country-industry level. *** p<0.01, ** p<0.05, * p<0.1.

### Robustness to other industry characteristics (Table 5)
- Tech intensity × Conflict c,t remains negative and significant when controlling for:
  - Physical capital intensity: -2.232*** (0.764)
  - Product complexity: -2.426*** (0.811)
  - Dependence on external finance: -3.367** (1.616)
  - Liquidity needs: -2.176*** (0.744)
  - Asset tangibility: -2.378*** (0.786)
  - Imports exposure: -2.741*** (1.043)
  - Exports exposure: -2.775*** (1.048)
- VA share remains negative and highly significant across specifications.

### Robustness to country-level heterogeneity (Table 6)
- Tech intensity × Conflict c,t remains negative and significant interacting with:
  - Real GDP per capita: -2.252*** (0.782)
  - Credit to private sector (% GDP): -1.835*** (0.878)
  - Chinn-Ito: -2.283*** (0.796)
  - Trade (% GDP): -2.274*** (0.806)
  - FDI interactions: -2.222*** (0.782)
  - Institutional quality: -2.255*** (0.778)
- Observations and R-squared vary by column (listed in source).

### Alternative proxies for technological intensity (Table 7)
- Differential estimated additional declines across proxies range from 0.7 to 6.6 percentage points:
  - RDI alternative I: estimated differential effect ≈ 0.7 percentage points (example: ISIC 29 vs. ISIC 27).
  - RDI alternative II: estimated differential effect ≈ 6.6 percentage points (largest).
  - Patent intensity, QC intensity, and OECD tech classification also yield significant negative effects (magnitudes differ due to scale differences across proxies).
- Selected interaction coefficients (Tech intensity_j × Conflict_c,t):
  - RDI alternative I: -39.786*** (14.435)
  - RDI alternative II: -276.509*** (97.108)
  - Patent intensity: -0.587** (0.279)
  - QC intensity: -0.214** (0.106)
  - Tech classification: -3.175** (1.533)
- Observations vary by proxy; R-squared reported per column.

### Subsamples and additional robustness (Table 8)
- Contemporaneous Tech intensity × Conflict c,t across specifications:
  - ≥10 years of data: -2.498*** (0.810)
  - Drop advanced economies: -2.330*** (0.799)
  - Countries with at least one year of conflict: -2.054*** (0.770)
  - Exclude US: -2.241*** (0.770)
  - Drop 5th and 95th percentile labor productivity extremes: -1.828*** (0.707)
  - Average VA share < 10 percent: -2.812*** (1.136)
  - Account for conflict size (1=mild, 2=intense): -1.949*** (0.675)
  - Deflate labor productivity growth by (1 + real exchange rate appreciation): -2.897*** (1.081)
- Observations and R-squared vary by column (listed in source).
- Randomization placebo tests: only 0.4 percent of 1000 placebo tests produced a t-statistic smaller than baseline, suggesting unobserved country-level shocks unlikely drive main result.

### Key exact statistics
- Additional decline in machinery vs. paper products: 1.5 percentage points (2.2 times the R&D intensity difference for those industries).
- Average annual industry labor productivity growth in sample: 4.1 percent.
- Table 4: VA share_c,j,t−1 = -1.498*** (0.087); Observations = 49,518; R-squared = 0.358.
- Selected Table 5 estimates: Tech intensity × Conflict c,t = -2.232*** (0.764); -2.426*** (0.811); -3.367** (1.616).
- Table 6 example: Tech intensity × Conflict c,t = -2.252*** (0.782).
- Table 7 percentile comparisons: 0.7; 6.6; 3.7; 5.2; 3.2 (estimated differential effects across proxies).

---

### 5. Interpretation, Implications, and Suggested Directions

### Interpretation
- Conflicts disproportionately reduce labor productivity growth in R&D-intensive industries; losses accumulate during protracted conflicts.
- Lack of short-term rebound suggests some R&D-enhancing investments are canceled or shelved rather than merely delayed, contributing to persistent output losses post-conflict.
- The R&D channel helps explain durable declines in GDP per capita following intrastate conflicts.

### Policy implications (paper’s scope)
- While specific conflict-prevention or containment policies are country-specific and beyond the paper’s scope, findings point to the need for policies supportive of innovative activities during conflicts.
- Suggested future research: identify which specific policies reduce economic costs of conflicts via the R&D channel in conflict-affected states.

*From IMF Working Paper — wpiea2024098-print-pdf (https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024098-print-pdf.pdf).*

### 1. Introduction ........................................................................................................

### 1. Introduction

### Background and motivation
- About one quarter of the world’s population lives in conflict-affected areas, with the number of conflicts reaching historical highs worldwide since the end of World War II, according to the UN.
- Violent conflicts impose dire humanitarian costs and are a major development challenge; literature typically finds a negative and enduring relationship between conflicts and economic performance, particularly for intrastate conflicts.
- Bi-directional associations between conflicts and poor economic performance (conflict → growth decline; poor growth → higher conflict risk) complicate causal inference using macro-data, creating “conflict traps.”

### Hypothesis and identification strategy
- Hypothesis: Intrastate conflicts have a disproportionate effect on industry growth in industries that rely more on R&D activities (higher technological intensity).
- Identification: Exploit within-country variation across 2-digit manufacturing industries based on technological intensity, in the spirit of Rajan and Zingales (1998).
- Benchmarking technological intensity: Industry-level R&D intensity measured using data on R&D expenditures of large, listed firms from the US (and robustness using OECD or patents/quality certificates).

### Data scope
- Industry coverage: 2-digit manufacturing industries (ISIC Rev. 3, 15-36).
- Country/sample coverage: 114 developing, emerging market, and advanced economies.
- Time period: 1980-2018.
- Conflict data: Years of intrastate conflicts from the Major Episodes of Political Violence (MEPV) database.
- Outcome variable: Industry labor productivity (used as equivalent to GDP per capita at macro level).

### Mechanisms: why R&D-intensive industries are more affected
- Characteristics of R&D investment making it vulnerable during conflicts:
  - Produces outcomes with a longer delay and is riskier than standard investment.
  - Requires larger scale and lumpy initial investment; firms may defer in times of uncertainty.
  - Involves highly specialized, low-redeployability, high-irreversibility assets.
- Institutional and market impairments during conflicts:
  - Reduced ability to enforce contracts, increased corruption, eroded trust → weaker intellectual property protection → appropriability problems → underinvestment in R&D.
  - Declines in consumer and investor confidence reduce demand for high-tech products and willingness to fund long-horizon projects.
  - Disruptions to infrastructure (electricity, internet) and to collaboration/knowledge spillovers, R&D cooperation, and outsourcing.

### Key empirical findings
- Conflicts lead to a decline in labor productivity growth, particularly in industries with higher technological intensity.
- Magnitude example: Labor productivity growth declines 1.5 percentage points more in machinery industry (industry at the 75th percentile of R&D intensity distribution) compared to paper products (industry at the 25th percentile of R&D intensity distribution) in each year with conflicts.
- Context: The average annual growth rate of labor productivity in the sample is about 4 percent.
- Pre-conflict trends: High-tech industries do not exhibit a different growth pattern in pre-conflict periods, supporting identification.
- Post-conflict dynamics: In the two-year window following conflicts, the additional decline in labor productivity growth in high-tech industries does not appear to persist, but there is no rebound either — lost productivity in high-tech industries during conflicts is not recovered in the short-term aftermath.
- Interpretation: Some productivity-enhancing R&D activities appear canceled or shelved for longer periods rather than merely delayed.

### Tests of alternative explanations and robustness
- Alternative channels tested and empirically addressed:
  - Reliance on physical capital intensity (conflicts hinder investment in physical capital).
  - Input variety and supply-chain disruptions.
  - Greater reliance on external finance, liquidity needs, and financing frictions affecting R&D.
  - Differential exposure to trade (imports/exports).
- Robustness checks:
  - Controls for country-level factors (macroeconomic, financial, institutional development; financial and trade openness).
  - Subsamples including emerging markets and developing economies.
  - Sample restricted to smaller 2-digit industries in manufacturing.
  - Randomization tests to rule out results driven by random country-level shocks.
- Result: The disproportionate effect of conflicts on growth in R&D-intensive industries remains similar across checks.

### Contribution to the literature
- Advances causal evidence in the conflict-growth nexus by using granular cross-country industry-level data and cross-industry variation in technological intensity.
- Identifies the R&D channel as a mechanism explaining durable declines in income post-conflict, given the role of innovation in long-term economic growth.
- Places findings in the context of literature showing long-lasting GDP per capita declines after intrastate conflicts (e.g., GDP per capita remains about 6 percentage points lower on average, even 10 years after intrastate conflicts; absence of conflicts since the 1970s would have made global GDP 12 percent higher in 2014, per cited studies).

### Organization of the paper
- Section 2: Data.
- Section 3: Empirical methodology.
- Section 4: Results.
- Section 5: Conclusion.

*From "1. Introduction", IMF WORKING PAPERS, INTERNATIONAL MONETARY FUND.*

### 2. Data

### 2. Data

### 2.1. Episodes of Conflicts
- Data source: Major Episodes of Political Violence (MEPV) database from Center for Systemic Peace.
- Event definition: systematic and sustained use of violence by organized groups where human toll exceeds 500 deaths.
- Temporal coverage: available until 2018.
- Focus: intrastate episodes of political violence (civil and ethnic violence and wars), to ensure events take place entirely on the country’s own soil.
- Coding:
  - For each country, a dummy variable = 1 for each year with intrastate conflicts.
  - Conflict “size” classification (separate test): assign variable 1 (2) for years whenever the score is below (above) the top quartile of the scores across the sample, and 0 when there is no conflict.
- Distributional note:
  - Total events in sample: 519 events of violent conflicts.
  - Coverage: conflicts cover about 19 percent of country-industry-year observations.
  - Geographic spread: 70, 279, 91 and 79 country-year observations with conflicts in Africa, Asia, Americas and Europe, respectively.
  - Vast majority of events (97 percent) occurred in emerging markets and developing economies.

### 2.2. Industry-level Data
- Source: UNIDO database (compiled largely based on industrial surveys).
- Industry coverage: 2-digit (ISIC Rev. 3) manufacturing industries (ISIC 15-36).
- Key variables:
  - Labor productivity = industry value added (in US dollars) / number of employees.
  - Growth rate of industry labor productivity = log difference, and multiplied by 100.
  - Industry value added share in total manufacturing (control) = ratio of industry value added to total value added in manufacturing (in percent).
- Data treatment:
  - Change in log labor productivity restricted to 100 and -100 percent (roughly corresponds to the 1st and 99th percentiles of the sample).
  - Confirmed that this restriction does not affect or drive results; results are also very similar if labor productivity growth is winsorized at the 1st and 99th percentile.
- Summary statistics: illustrated in the Appendix (Table A.1).

### 2.3. Industry Technological Intensity
- Objective: explore effect of conflicts on labor productivity growth across industries with different inherent technological intensity.
- Identification strategy: exploit cross-industry variation in technological intensity using benchmarked proxies (in the spirit of Rajan and Zingales (1998)).
- Baseline proxy (RDI baseline):
  - Measure: intensity of R&D expenditures in overall capital expenditures.
  - Source: Igan et al. (2022), based on publicly listed firms in the US from the Compustat database.
  - Construction: average firm-level annual ratios over 1980–1999; median across firms within each industry adopted as industry proxy.
  - Sample statistics: mean value = 0.97; standard deviation = 1.34 (see Table A.1).
  - 25th percentile industry: paper products (ISIC 21); 75th percentile industry: machinery (ISIC 29).
- Rationale for US-based benchmarking:
  - Highly developed financial markets, strong legal and enforcement institutions, large technology consumption market, and accessibility of capital markets reduce financing- and institution-driven distortions in R&D measures.
  - Ensures industry technological intensity is exogenous to country-specific developments or shocks.
- Robustness proxies:
  - RDI alternative I: ratio of R&D expenditures to sales for firms in the US (Brown et al. 2017); median across firms, summed over the 1990s.
  - RDI alternative II: share of industry R&D expenditure in output in the OECD, averaged over 1990-2009 (Markus et al. 2019).
  - Patent intensity: patent counts divided by value added in million US dollars (NBER Patent Database, USPTO applications 1980-1986), following Aghion et al. (2015).
  - QC (quality certificate) intensity: average share of firms holding quality certificates in each industry over 2006–2011 (World Bank Enterprise Surveys, Paunov 2016).
- Concordance across proxies:
  - Proxies are positively correlated (including RDI baseline with RDI alternative I and RDI alternative II).
  - Industries above median and quartile classifications mostly overlap across proxies.
- Categorization (Galindo-Rueda and Verger 2016):
  - Assign industry technological intensity variable: 3 for high-tech (ISIC 24, 29-35), 2 for medium-tech (ISIC 23, 25-28), 1 for low-tech.
  - Note: classification very similar to RDI baseline (7 out of 8 high-tech industries are among top 8 in RDI baseline).

- Selected RDI baseline and alternative values from Table 1 (industry, ISIC, RDI baseline, RDI alternative I, RDI alternative II, Patent intensity, QC intensity):
  - Food and beverages 15 — 0.180, 0.0065, 0.002, 0.2, 23.1
  - Tobacco products 16 — 0.402, 0.0045, 0.004, 0.0, 41.1
  - Textiles 17 — 0.349, 0.0127, 0.005, 0.4, 16.2
  - Wearing apparel, fur 18 — 0.000, 0.0000, 0.002, 0.4, 11.1
  - Leather, leather products, footwear 19 — 0.554, 0.0069, 0.004, 0.4, 9.4
  - Wood products (excl. furniture) 20 — 0.164, 0.0048, 0.002, 0.1, 12.6
  - Paper and paper products 21 — 0.266, 0.0097, 0.003, 0.2, 24.1
  - Printing and publishing 22 — 0.602, 0.0082, 0.001, 0.2, 13.1
  - Coke, refined petroleum, nuclear fuel 23 — 0.062, 0.0057, 0.004, [blank], [blank]
  - Chemicals and chemical products 24 — 5.456, 0.2748, 0.035, 6.4, 35.1
  - Rubber and plastics products 25 — 0.368, 0.0139, 0.009, 1.4, 31.0
  - Non-metallic mineral products 26 — 0.337, 0.0116, 0.005, 0.6, 20.5
  - Basic metals 27 — 0.168, 0.0069, 0.005, 0.2, 28.6
  - Fabricated metal products 28 — 0.351, 0.0096, 0.004, 0.9, 27.4
  - Machinery and equipment n.e.c. 29 — 0.926, 0.0232, 0.017, 10.0, 31.8
  - Office, accounting, computing machinery 30 — 2.845, 0.1205, 0.059, 19.0, 36.8
  - Electrical machinery and apparatus 31 — 1.245, 0.0318, 0.028, 11.3, 46.6
  - Radio, TV, communication equipment 32 — 2.353, 0.1054, 0.065, 25.5, 53.2
  - Medical, precision, optical instruments 33 — 3.168, 0.1053, 0.047, 18.7, 42.9
  - Motor vehicles, trailers, semi-trailers 34 — 0.441, 0.0166, 0.018, 1.4, 44.6
  - Other transport equipment 35 — 0.614, 0.0206, 0.031, 0.3, 43.2
  - Furniture; manufacturing n.e.c. 36 — 0.511, 0.0177, 0.004, 2.3, 10.4

### 2.4. Other Industry Characteristics
- Purpose: test whether technological intensity proxies capture other industry characteristics.
- Firm-level-derived measures (Igan et al. 2022), averaged over 1980-1999 and median across firms within each industry:
  - Dependence on external finance: share of firm’s capital expenditures not financed with cash flow from operations (Rajan and Zingales 1998).
  - Liquidity needs: ratio of a firm’s total inventories to annual sales (Raddatz 2006).
  - Asset tangibility: share of net property, plant and equipment in total book-value of firm assets (Braun and Larrain 2005).
- Other industry-level measures:
  - Physical capital intensity: total real capital stock / total value added in each industry from NBER-CES Manufacturing Industry Database, averaged over 1980-1999.
  - Product complexity (input concentration): industry Herfindahl index of intermediate input use (multiplied by minus 1) computed from US 1992 I-O Use Table (Levchenko 2007).
- Trade exposure proxies:
  - Source: OECD STAN Bilateral Trade Database by Industry and end-use category (BTDIxE).
  - Construction: sum US dollar values of industry-level exports (and imports) over the 1990s; industry’s share in total manufacturing exports (or imports) in OECD countries.
  - Note: for broader-than-2-digit ISIC classifications, measures equally divided to obtain 2-digit ISIC values (for ISIC 15-16, ISIC 17-19, and ISIC 21-22).
  - Normalization: industry exports (or imports) normalized with total exports (or imports) in manufacturing.
- Selected values from Table 3 (industry, ISIC, Physical capital intensity, Product complexity, Dependence on external finance, Liquidity needs, Asset tangibility, Imports, Exports):
  - Food and beverages 15 — 1.466, -0.111, -0.309, 0.102, 0.363, 0.037, 0.035
  - Tobacco products 16 — 1.047, -0.118, -2.897, 0.252, 0.219, 0.037, 0.035
  - Textiles 17 — 1.287, -0.121, -0.157, 0.178, 0.329, 0.030, 0.020
  - Wearing apparel, fur 18 — 0.440, -0.090, -0.298, 0.213, 0.116, 0.030, 0.020
  - Chemicals and chemical products 24 — 1.768, -0.069, 3.868, 0.141, 0.206, 0.110, 0.120
  - Machinery and equipment n.e.c. 29 — 0.995, -0.039, -0.431, 0.203, 0.216, 0.098, 0.125
  - Office, accounting, computing machinery 30 — 0.752, -0.086, 1.056, 0.190, 0.133, 0.064, 0.051
  - Motor vehicles, trailers, semi-trailers 34 — 1.365, -0.049, -0.140, 0.142, 0.270, 0.126, 0.139
  - Furniture; manufacturing n.e.c. 36 — 0.764, -0.041, -0.337, 0.176, 0.204, 0.039, 0.030

### 2.5. Country-level Data
- Purpose: robustness checks controlling for macroeconomic, financial and institutional variables.
- Macroeconomic and financial variables (World Bank WDI):
  - Real GDP per capita (constant in US dollars) — proxy for economic development.
  - Credit to the private sector (percent of GDP) — proxy for financial development.
  - Trade (percent of GDP) — proxy for trade openness.
- Financial openness:
  - Chinn and Ito (2006) index — de jure (regulatory) measure of capital account openness, scaled between 0 and 1 (higher = more open).
  - Alternate: FDI (stock of liabilities, percent of GDP) from Milesi-Ferretti (2022) as a de facto measure of financial globalization.
- Institutional quality:
  - Constraints on executives index from Polity V (Center for Systemic Peace), ranges between 0 and 7 (higher = stronger institutional constraints).
- Real exchange rate appreciation:
  - Data: nominal exchange rates (local currency vis-a-vis USD) from WDI.
  - CPI and PPI inflation from World Bank inflation database (constructed by Ha et al. 2023).

### 2.6. Sample
- Time span: all available data starting from 1980, main sample period 1980-2018.
- Coverage: 2-digit manufacturing industries (ISIC 15-36) over 1980-2018.
- Country coverage: 114 countries, with 21 being advanced economies.
- Data treatment: exclude country-industry pairs with less than 5 years of available data (step does not affect results).
- Conflicts summary:
  - 519 events of violent conflicts in the sample.
  - Conflicts represent about 19 percent of country-industry-year observations.
  - Figure 1 reports number of countries with intrastate conflicts each year over sample period.

*Source: IMF Working Paper — Section 2. Data (wpiea2024098-print-pdf).*

### 3. Methodology

### 3. Methodology

### Goal and conceptual approach
- Goal: examine the heterogeneous effects of conflicts on growth based on variation in industries’ technological intensity.
- Empirical framework: difference-in-differences approach pioneered by Rajan and Zingales (1998), exploiting differential impacts of a country-level shock (conflict) on industries with inherently different technological intensity (R&D reliance).
- Core hypothesis: if conflicts affect innovative activities, the impact on growth should be more pronounced in industries with relatively higher technological intensity.

### Identification assumptions
- Two implicit assumptions (aligned with Rajan and Zingales (1998)):
  - Some industries inherently and persistently rely more on R&D activities than others, mainly driven by production processes.
  - These innate differences carry over countries, so an industry’s technological intensity calculated using US data serves as a proxy for its technological intensity in other countries.

### Baseline econometric specification (equation (1))
- Specification:
  Δlog(퐿푃푐,푗,푡) = 훼 푉퐴 푠ℎ푎푟푒푐,푗,푡−1
  + ∑훽푟 푇푒푐ℎ 푖푛푡푒푛푠푖푡푦푗 × 퐶표푛푓푙푖푐푡푐,푡−푟
  2
  푟=−2
  + 휃푐,푡 + 휃푐,푗 + 휃푗,푡 + 휖푐,푗,푡
- Notation:
  - 푐, 푗, 푡 = countries, 2-digit manufacturing industries, and years.
  - Dependent variable Δlog(퐿푃푐,푗,푡) = growth rate of industry labor productivity (in percentage points).
  - 푇푒푐ℎ 푖푛푡푒푛푠푖푡푦푗 = proxy for industry’s technological intensity (enables comparison across 2-digit manufacturing industries).
  - 퐶표푛푓푙푖푐푡푐,푡−푟 = dummy variable indicating the years of conflicts.
  - Main variables of interest: interaction terms 푇푒푐ℎ 푖푛푡푒푛푠푖푡푦푗 × 퐶표푛푓푙푖푐푡푐,푡−푟, for r = −2, −1, 0, 1, 2 (five-year window around conflicts).
- Interpretation of 훽푟:
  - 훽−2 and 훽−1: assess pre-conflict (two-year) patterns in labor productivity growth.
  - 훽0: captures effect in the years of conflicts.
  - 훽1 and 훽2: capture aftermath dynamics (persistence or rebound).

### Pre-trends and identification strategy
- Inclusion of forward values of conflict dummy (r = −1, −2) explicitly tests pre-trends: identification requires no distinct growth pattern in high-tech industries during pre-conflict years.
- Expectation:
  - 훽−2 and 훽−1 to be statistically insignificant.
  - If conflicts adversely and disproportionately impact high-tech industries: 훽0 expected negative and statistically significant.
  - If effect persists: 훽1 and 훽2 expected negative; if positive, suggests a post-conflict growth rebound.

### Fixed effects, controls, and standard errors
- Fixed effects included:
  - Country-year (휃푐,푡): controls for time-variant country-wide factors (economic, political, institutional developments, country-specific annual shocks including demand or supply shocks common across industries).
  - Country-industry (휃푐,푗): controls for persistent, country-specific industry characteristics (e.g., if high-tech industries systematically grow differently in some countries).
  - Industry-year (휃푗,푡): controls for industry-specific global developments, trends, growth opportunities, or shocks at annual frequency.
- Additional control: lagged value of industry value added share in total manufacturing (푉퐴 푠ℎ푎푟푒푐,푗,푡−1).
- Standard errors: clustered at the country-industry level.
  - Robustness note (footnote 14): results remain virtually the same if standard errors are (i) not clustered but robust to heteroskedasticity, (ii) three-way clustered (country-industry, country-year, industry-year), (iii) two-way clustered with any combinations, or (iv) one-way clustered at country, industry or year levels.

### Advantages of industry-level approach and reverse causality
- Industry-level data mitigate reverse causality concerns: changes in growth of a R&D intensive 2-digit manufacturing industry are unlikely to significantly influence the likelihood of conflict in a country, unlike macro-data.

### Extensions: testing alternative industry characteristics (equation (2))
- Extended specification:
  Δlog(퐿푃푐,푗,푡) = 훼 푉퐴 푠ℎ푎푟푒푐,푗,푡−1
  + ∑훽푟 푇푒푐ℎ 푖푛푡푒푛푠푖푡푦푗 × 퐶표푛푓푙푖푐푡푐,푡−푟
  2
  푟=−2
  + ∑휇푟 퐼푗 × 퐶표푛푓푙푖푐푡푐,푡−푟
  2
  푟=−2
  + 휃푐,푡 + 휃푐,푗 + 휃푗,푡 + 휖푐,푖,푡
- 퐼푗 represents other industry characteristics: physical capital intensity, input concentration, dependence on external finance, liquidity needs, asset tangibility, exposure to imports and exports.
- Purpose: examine whether technological intensity proxies for other industry features and to assess heterogeneous conflict impacts along these other dimensions.

### Extensions: heterogeneous country-level effects (equation (3))
- Specification:
  Δlog(퐿푃푐,푗,푡) = 훼 푉퐴 푠ℎ푎푟푒푐,푗,푡−1
  + ∑훽푟 푇푒푐ℎ 푖푛푡푒푛푠푖푡푦푗 × 퐶표푛푓푙푖푐푡푐,푡−푟
  2
  푟=−2
  + ∑훾푟 푇푒푐ℎ 푖푛푡푒푛푠푖푡푦푗 × 𝑋푐,푡−푟
  2
  푟=−2
  + 휃푐,푡 + 휃푐,푗 + 휃푗,푡 + 휖푐,푖,푡
- 𝑋푐,푡−푟 are proxies for country-level factors: economic and financial development, financial openness, trade openness, FDI, institutional quality.
- Purpose: test whether country-level variables have effects that vary with industry technological intensity and whether conflicts might proxy for other country-level factors.

### Expectations from extensions
- If (i) industry technological intensity is not a proxy for other industry features, and (ii) conflicts do not serve as a proxy for other country-level factors, then results from specifications (2) and (3) should be similar to the baseline specification (1).

### Preview of empirical findings (from Section 4)
- Baseline results (Table 4 and Figure 2, described):
  - Labor productivity growth in high-tech industries does not exhibit a distinct pre-conflict pattern (supporting identification).
  - The year t coefficient is negative and statistically significant at the 1 percent level, suggesting labor productivity declines in the years of conflicts, particularly in industries with higher technological intensity.
  - Quantitative example: the two-digit industries at the 25th and 75th percentiles of R&D intensity distribution are paper products (ISIC 21) and machinery (ISIC 29). The proxy for technological intensity is 0.266 for paper products, and 0.926 for machinery (Table 1, RDI baseline). The coefficient estimate of [continuation in results section].

*IMF Working Paper — Section 3. Methodology (extracted content). *

### 2.2 in the years of conflicts suggests that labor productivity growth declines by an additional 1.5

### wpiea2024098-print-pdf - 2.2 in the years of conflicts suggests that labor productivity growth declines by an additional 1.5

### Key findings
- Conflicts disproportionately reduce labor productivity growth in industries with higher technological intensity; the estimated additional decline in machinery relative to paper products is 1.5 percentage points (2.2 times the difference in R&D intensity measures for those two industries).
- The average annual growth rate of industry labor productivity in the sample is 4.1 percent.
- The Rajan and Zingales (1998) style interaction estimates quantify the differential effect across industries with relatively higher technological intensity, not the overall impact on aggregate industry growth.
- The disproportionate negative effect on high-tech industries occurs during each year of protracted conflicts, accumulating larger losses for durable conflicts.
- No evidence of a short-term rebound: the two-year window following conflicts does not show significantly higher growth that would make up for losses in high-tech industries; some R&D-enhancing investments appear canceled or shelved for longer periods rather than merely delayed.

### Baseline empirical results (Table 4)
- Tech intensity × Conflict c,t (key contemporaneous impact):
  - Column (1): -1.325** (0.543)
  - Column (2)/(3): -2.236*** (0.770)  (reported in table as a stronger negative effect in expanded specifications)
- Lagged and lead interaction coefficients (selected):
  - Tech intensity × Conflict c,t+2: 0.705 (0.585); 0.732 (0.584); 0.764 (0.583)
  - Tech intensity × Conflict c,t+1: -0.360 (0.592); 0.669 (0.793); 0.609 (0.802)
  - Tech intensity × Conflict c,t−1: 0.728 (0.916)
  - Tech intensity × Conflict c,t−2: 0.590 (0.661)
- Industry initial value-added share:
  - VA share c,j,t−1: -1.498*** (0.087) (smaller industries tend to exhibit higher labor productivity growth)
- Estimation details:
  - Country-year F.E.: Yes
  - Country-industry F.E.: Yes
  - Industry-year F.E.: Yes
  - Observations: 49,518
  - R-squared: 0.358
- Notes: Standard errors in parentheses are clustered at the country-industry level. *** p<0.01, ** p<0.05, * p<0.1.

### Robustness checks: other industry characteristics (Table 5)
- The baseline negative contemporaneous effect of conflicts on high-tech industries is robust to controlling for:
  - Physical capital intensity (column 1): Tech intensity × Conflict c,t = -2.232*** (0.764)
  - Product complexity (column 2): Tech intensity × Conflict c,t = -2.426*** (0.811)
  - Dependence on external finance (column 3): Tech intensity × Conflict c,t = -3.367** (1.616)
  - Liquidity needs (column 4): Tech intensity × Conflict c,t = -2.176*** (0.744)
  - Asset tangibility (column 5): Tech intensity × Conflict c,t = -2.378*** (0.786)
  - Imports exposure (column 6): Tech intensity × Conflict c,t = -2.741*** (1.043)
  - Exports exposure (column 7): Tech intensity × Conflict c,t = -2.775*** (1.048)
- Selected additional findings:
  - Some evidence that industries with greater reliance on external finance experience a disproportionately negative effect with one-year delay, but also evidence of short-term recovery via this channel (column 3).
  - Industry initial value-added share remains negative and highly significant in all specifications: VA share c,j,t−1 ≈ -1.497*** (0.087).
- Estimation details:
  - Observations: 49,518
  - R-squared: 0.358
- Notes: I stands for the industry-level characteristic interacted with conflict (e.g., physical capital intensity, product complexity, dependence on external finance, liquidity needs, asset tangibility, import share, export share). Standard errors clustered at country-industry level. *** p<0.01, ** p<0.05, * p<0.1.

### Robustness checks: other country-level factors (Table 6)
- The baseline differential effect remains when interacting technological intensity with country-level variables:
  - Economic development (real GDP per capita) specification: Tech intensity × Conflict c,t = -2.252*** (0.782)
  - Financial development (credit to private sector as percent of GDP): Tech intensity × Conflict c,t = -1.835*** (0.878)
  - Financial openness (Chinn-Ito): Tech intensity × Conflict c,t = -2.283*** (0.796)
  - Trade openness (trade as percent of GDP): Tech intensity × Conflict c,t = -2.274*** (0.806)
  - FDI interactions: Tech intensity × Conflict c,t = -2.222*** (0.782)
  - Institutional quality (constraints on executives): Tech intensity × Conflict c,t = -2.255*** (0.778)
- Estimation details:
  - Observations vary by column: 47,560; 33,260; 47,516; 46,746; 48,976; 49,518
  - R-squared varies by column: 0.355; 0.358; 0.359; 0.367; 0.352; 0.358
- Notes: X stands for the country-level variable in each column. Standard errors clustered at country-industry level. *** p<0.01, ** p<0.05, * p<0.1.

### Alternative proxies for technological intensity (Table 7)
- Results are robust when technological intensity is proxied by:
  - R&D intensity as percent of sales (RDI alternative I)
  - R&D intensity from OECD data (RDI alternative II)
  - Patenting intensity
  - Quality certificate intensity
  - OECD classification of high-, medium-, and low-tech industries
- Magnitude varies across proxies:
  - Using RDI alternative I: the industry at the 75th percentile (ISIC 29) declines about 0.7 percentage points more in the years of conflicts compared to the industry at the 25th percentile (ISIC 27).
  - Using R&D expenditure from OECD (column 3): estimated differential impact across 25th and 75th percentiles is 6.6 percentage points (largest among measures).
  - OECD classification-based estimate (last column) yields a distinct estimated coefficient (reported in Table 7; magnitudes differ across proxies).

### Interpretation and implications
- The negative and statistically significant contemporaneous coefficients for Tech intensity × Conflict indicate that high-tech (R&D-intensive) industries suffer larger productivity losses during conflict years relative to lower-tech industries.
- Lack of short-term rebound implies permanent or prolonged losses in productivity-enhancing investments in high-tech industries following conflicts.
- The effect is not driven by alternative industry characteristics (physical capital reliance, product complexity, financing dependence, liquidity needs, asset tangibility) or by a range of country-level factors (economic development, financial development, openness, trade, FDI, institutional quality).
- Smaller industries tend to exhibit higher productivity growth, but the baseline differential impact of conflicts on high-tech industries is robust to controlling for industry size.

### Key exact statistics (selected)
- Additional decline in machinery vs. paper products: 1.5 percentage points (2.2 times the difference in R&D intensity measures)
- Average annual growth rate of industry labor productivity in sample: 4.1 percent
- Table 4: VA share c,j,t−1 = -1.498*** (0.087)
- Table 4 estimations: Observations = 49,518; R-squared = 0.358
- Table 5 examples: Tech intensity × Conflict c,t = -2.232*** (0.764); -2.426*** (0.811); -3.367** (1.616)
- Table 6 example: Tech intensity × Conflict c,t = -2.252*** (0.782)
- Table 7 percentile comparisons: 0.7 percentage points (RDI alternative I); 6.6 percentage points (RDI alternative II)

*Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024098-print-pdf.pdf*

### 3.2 percentage points larger impact on labor productivity growth in high-tech industries relative to

### wpiea2024098-print-pdf - 3.2 percentage points larger impact on labor productivity growth in high-tech industries relative to

### Key findings on technological-intensity differential effects
- The estimated additional decline in labor productivity in high-tech industries relative to less technologically intensive industries ranges from 0.7 to 6.6 percentage points (reported as "Estimated differential effect 0.7 6.6 3.7 5.2 3.2" across proxies).
- Despite sensitivity to the specific proxy (range from 0.7 to 6.6 percentage points), the coefficient estimates remain statistically significant and economically sizable across proxies.
- Comparing coefficient magnitudes directly is misleading because the scales of proxies for technological intensity are not comparable; the paper focuses on the additional decline in growth in high-tech industries relative to less intensive industries based on each measure.

### Table 7: Other proxies of technological intensity (selected estimates and statistics)
- Interaction coefficients (Tech intensity_j × Conflict_c,t) at t (contemporaneous):
  - RDI alternative I: -39.786*** (14.435)
  - RDI alternative II: -276.509*** (97.108)
  - Patent intensity: -0.587** (0.279)
  - QC intensity: -0.214** (0.106)
  - Tech classification: -3.175** (1.533)
- Lead and lag interaction coefficients reported for t+2, t+1, t−1, t−2 with standard errors in parentheses (see Table 7 for full matrix).
- VA share c,j,t−1: -1.498*** (0.087) (consistent across columns reported)
- Estimated differential effect (by proxy): 0.7; 6.6; 3.7; 5.2; 3.2
- Fixed effects included: Country-year F.E., Country-industry F.E., Industry-year F.E. (Yes for all columns)
- Observations: 49,518; 49,518; 47,550; 47,550; 49,518 (by column)
- R-squared: 0.358; 0.358; 0.373; 0.373; 0.358 (by column)
- Notes: Results based on equation 1. Proxies for technological intensity described in Section 2. Standard errors clustered at country-industry level. *** p<0.01, ** p<0.05, * p<0.1.

### Additional tests (Section 4.5) — robustness and subsamples (Table 8)
- Subsample and robustness exercises summarized by columns (1)–(8); key reported estimates for contemporaneous interaction (Tech intensity_j × Conflict_c,t):
  - Column (1) [country-industry pairs with ≥10 years of data]: -2.498*** (0.810)
  - Column (2) [drops advanced economies]: -2.330*** (0.799)
  - Column (3) [countries with at least one year of conflict]: -2.054*** (0.770)
  - Column (4) [excludes the US]: -2.241*** (0.770)
  - Column (5) [drops 5th and 95th percentile labor productivity extremes]: -1.828*** (0.707)
  - Column (6) [average VA share < 10 percent]: -2.812*** (1.136)
  - Column (7) [accounts for size of conflicts (1=mild, 2=intense)]: -1.949*** (0.675)
  - Column (8) [deflates labor productivity growth with (1 + appreciation in real exchange rate)]: -2.897*** (1.081)
- Lead and lag interaction coefficients for these specifications are reported with standard errors in Table 8 (e.g., t+2 and t+1 positive but smaller and statistically insignificant in most columns; t−1 and t−2 generally small and insignificant).
- VA share coefficients in Table 8 vary across specifications (examples):
  - Column (1): -1.492*** (0.087)
  - Column (3): -1.776*** (0.153)
  - Column (6): -2.549*** (0.165)
- Observations by column: 47,464; 35,747; 17,402; 48,792; 44,963; 43,058; 49,518; 49,252
- R-squared by column: 0.360; 0.351; 0.421; 0.358; 0.382; 0.368; 0.358; 0.391
- Notes clarify each column’s sample selection and that standard errors are clustered at country-industry level. *** p<0.01, ** p<0.05, * p<0.1.

### Tests for reverse causality and confounding shocks
- Reverse causality addressed by focusing on smaller two-digit manufacturing industries (average VA share < 10 percent); the adverse impact on labor productivity growth in high-tech industries remains pronounced (see Column 6).
- Magnitude of conflicts examined by coding conflict variable as 1 for mild and 2 for more intense episodes; results point to a larger effect for more intense episodes (Column 7).
- Real exchange rate movements accounted for by deflating labor productivity growth with (one plus) the rate of appreciation of the real exchange rate (Rodrik (2013) approach); results remain similar (Column 8).
- Randomization (placebo) tests:
  - Procedure: keep country-level average annual probability of conflicts unchanged; randomly assign placebo conflict dummy 1000 times; estimate full specification (1) each time.
  - Result: only 0.4 percent of 1000 placebo tests generate a t-statistic smaller than the baseline t-statistic for the coefficient at year t (β0), indicating that unobserved country-level shocks are unlikely to drive the main result.

### Conclusion and policy implications
- Conflicts are associated with a long-lasting drag on economic output, and this paper identifies a channel through which conflicts can reduce output: by disproportionately reducing innovative activities.
- Main empirical finding: conflicts lead to a disproportionate decline in labor productivity growth in industries with greater technological intensity; the differential effect in high-tech industries is estimated to be large.
- The additional labor productivity loss in high-tech industries during years of conflict is not offset in the post-conflict period.
- Implication: Given the link between innovative activities and long-term growth, these differential impacts help explain durable income losses after conflicts.
- Policy note: Specific policies to prevent or contain violent conflicts are country-specific and beyond the paper’s scope. However, findings point to the need for policies supportive of innovative activities during conflicts.
- Suggested future research: Explore which specific policies help reduce economic costs of conflicts through the R&D channel in conflict-affected states.

*Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024098-print-pdf.pdf*

### References

### References

### Bibliographic sources
- Aghion, Philippe, and Peter Howitt. “A model of growth through creative destruction.” Econometrica, 60(2), 323–351 (1992).
- Aghion, Philippe, Peter Howitt, and Susanne Prantl. "Patent rights, product market reforms, and innovation." Journal of Economic Growth 20 (2015): 223-262.
- Aghion, Philippe, and Xavier Jaravel. "Knowledge spillovers, innovation and growth." The Economic Journal 125, no. 583 (2015): 533-573.
- Ahn, JaeBin, Romain A. Duval, and Can Sever. Macroeconomic Policy, Product Market Competition, and Growth: The Intangible Investment Channel. No. 2020/025. International Monetary Fund, 2020.
- Alderman, Harold, John Hoddinott, and Bill Kinsey. "Long term consequences of early childhood malnutrition." Oxford Economic Papers 58, no. 3 (2006): 450-474.
- Anokhin, Sergey, and William S. Schulze. "Entrepreneurship, innovation, and corruption." Journal of Business Venturing 24, no. 5 (2009): 465-476.
- Bhattacharya, Utpal, Po-Hsuan Hsu, Xuan Tian, and Yan Xu. "What affects innovation more: policy or policy uncertainty?." Journal of Financial and Quantitative Analysis 52, no. 5 (2017): 1869-1901.
- Bircan, Çağatay, Tilman Brück, and Marc Vothknecht. "Violent conflict and inequality." Oxford Development Studies 45, no. 2 (2017): 125-144.
- Blattman, Christopher, and Edward Miguel. "Civil war." Journal of Economic Literature 48, no. 1 (2010): 3-57.
- Blomberg, S. Brock, and Gregory D. Hess. "The temporal links between conflict and economic activity." Journal of Conflict Resolution 46, no. 1 (2002): 74-90.
- Blomberg, S. Brock, Gregory D. Hess, and Siddharth Thacker. "On the conflict–poverty nexus." Economics & Politics 18, no. 3 (2006): 237-267.
- Bloom, Nicholas. "The impact of uncertainty shocks." Econometrica 77, no. 3 (2009): 623-685.
- Braun, Matias, and Borja Larrain. "Finance and the business cycle: international, inter‐industry evidence." The Journal of Finance 60, no. 3 (2005): 1097-1128.
- Brodeur, Abel. "The effect of terrorism on employment and consumer sentiment: Evidence from successful and failed terror attacks." American Economic Journal: Applied Economics 10, no. 4 (2018): 246-282.
- Brown, James R., Gustav Martinsson, and Bruce C. Petersen. "What promotes R&D? Comparative evidence from around the world." Research Policy 46, no. 2 (2017): 447-462.
- Bundervoet, Tom, Philip Verwimp, and Richard Akresh. "Health and civil war in rural Burundi." Journal of Human Resources 44, no. 2 (2009): 536-563.
- Center for Systemic Peace. Major Episodes of Political Violence 1946-2018 Database.
- Center for Systemic Peace. Polity V Database.
- Cerra, Valerie, and Sweta Chaman Saxena. "Growth dynamics: the myth of economic recovery." American Economic Review 98, no. 1 (2008): 439-457.
- Chupilkin, M and Z Kóczán (2022), “The Economic Consequences of War: Estimates Using Synthetic Controls”, EBRD Working Paper.
- Chamarbagwala, Rubiana, and Hilcías E. Morán. "The human capital consequences of civil war: Evidence from Guatemala." Journal of Development Economics 94, no. 1 (2011): 41-61.
- Chen, Min, and Yonghong Tang. "The shock of armed conflict on research and development investment: empirical evidence from cross-country data." Technology Analysis & Strategic Management (2023): 1-14.
- Chinn, Menzie D., and Hiro Ito. "What matters for financial development? Capital controls, institutions, and interactions." Journal of Development Economics 81, no. 1 (2006): 163-192.
- Ciccone, Antonio, and Elias Papaioannou. "Human capital, the structure of production, and growth." The Review of Economics and Statistics 91, no. 1 (2009): 66-82.
- Collier, Paul, and Anke Hoeffler. "On economic causes of civil war." Oxford Economic Papers 50, no. 4 (1998): 563-573.
- Collier, Paul. "On the economic consequences of civil war." Oxford Economic Papers 51, no. 1 (1999): 168-183.
- Collier, Paul, V. L. Elliott, Håvard Hegre, Anke Hoeffler, Marta Reynal-Querol, and Nicholas Sambanis. "Breaking the Conflict Trap: Civil War and Development Policy." World Bank Publications-Books (2003).
- Collier, Paul, and Marguerite Duponchel. "The economic legacy of civil war: firm-level evidence from Sierra Leone." Journal of Conflict Resolution 57, no. 1 (2013): 65-88.
- Corral, Paul, Alexander Irwin, Nandini Krishnan, and Daniel Gerszon Mahler. Fragility and conflict: On the front lines of the fight against poverty. World Bank Publications, 2020.
- Crost, Benjamin, Joseph Felter, and Patrick Johnston. "Aid under fire: Development projects and civil conflict." American Economic Review 104, no. 6 (2014): 1833-1856.
- Davis, Donald R., and David E. Weinstein. "Bones, bombs, and break points: the geography of economic activity." American Economic Review 92, no. 5 (2002): 1269-1289.
- De Groot, Olaf J., Carlos Bozzoli, Anousheh Alamir, and Tilman Brück. "The global economic burden of violent conflict." Journal of Peace Research 59, no. 2 (2022): 259-276.
- Dixit, Avinash K., and Robert S. Pindyck. Investment under uncertainty. Princeton University Press, 1994.
- Fang, Xiangming, Siddharth Kothari, Cameron McLoughlin, and Mustafa Yenice. "The Economic Consequences of Conflict in Sub-Saharan Africa." IMF Working Papers 2020, no. 221 (2020).
- Fearon, James D., and David D. Laitin. "Ethnicity, insurgency, and civil war." American Political Science review 97, no. 1 (2003): 75-90.
- Fich, Eliezer M., Tung Nguyen, and Dimitris Petmezas. "Uncertainty and corporate innovation: evidence from terrorist attacks." Available at SSRN 3394564 (2020).
- Galindo-Rueda, Fernando, and Fabien Verger. "OECD taxonomy of economic activities based on R&D intensity." OECD Working Papers (2016).
- Gates, Scott, Håvard Hegre, Håvard Mokleiv Nygård, and Håvard Strand. "Development consequences of armed conflict." World Development 40, no. 9 (2012): 1713-1722.
- Ghobarah, Hazem Adam, Paul Huth, and Bruce Russett. "Civil wars kill and maim people—long after the shooting stops." American Political Science Review 97, no. 2 (2003): 189-202.
- Gross, Daniel P., and Bhaven N. Sampat. "America, jump-started: World War II R&D and the takeoff of the US innovation system." American Economic Review 113, no. 12 (2023): 3323-3356.
- Grossman, Gene M., and Elhanan Helpman. Innovation and growth in the global economy. MIT Press, 1993.
- Ha, Jongrim, M. Ayhan Kose, and Franziska Ohnsorge. "One-stop source: A global database of inflation." Journal of International Money and Finance (2023): 102896.
- Hall, Bronwyn H., and Josh Lerner. "The financing of R&D and innovation." In Handbook of the Economics of Innovation, vol. 1, pp. 609-639. North-Holland, 2010.
- Hardy, Bryan, and Can Sever. "Financial crises and innovation." European Economic Review 138 (2021): 103856.
- Hoeffler, Anke, and Marta Reynal-Querol. "Measuring the costs of conflict." Washington, DC: World Bank (2003).
- Humphreys, Macartan. "Economics and violent conflict." Cambridge, MA 31 (2003).
- Igan, Deniz, Alexandre R. Lauwers, and Damien Puy. Capital flows and institutions. Bank for International Settlements, Monetary and Economic Department, 2022.
- Ilyina, Anna, and Roberto Samaniego. "Technology and financial development." Journal of Money, Credit and Banking 43, no. 5 (2011): 899-921.
- Imai, Kosuke, and Jeremy M. Weinstein. "Measuring the economic impact of civil war." CID Working Paper Series (2000).
- Institute for Economics & Peace. Global Peace Index 2023: Measuring Peace in a Complex World, Sydney, June 2023. Available from: http://visionofhumanity.org/resources.
- Iqbal, Zaryab, and Christopher Zorn. "Violent conflict and the spread of HIV/AIDS in Africa." The Journal of Politics 72, no. 1 (2010): 149-162.
- Jordà, Òscar. "Estimation and inference of impulse responses by local projections." American Economic Review 95, no. 1 (2005): 161-182.
- Kim, Hyunseob, and Howard Kung. "The asset redeployability channel: How uncertainty affects corporate investment." The Review of Financial Studies 30, no. 1 (2017): 245-280.
- Knight, Malcolm, Norman Loayza, and Delano Villanueva. "The peace dividend: military spending cuts and economic growth." IMF Staff Papers 43, no. 1 (1996): 1-37.
- Kogan, Leonid, Dimitris Papanikolaou, Amit Seru, and Noah Stoffman. "Technological innovation, resource allocation, and growth." The Quarterly Journal of Economics 132, no. 2 (2017): 665-712.
- Lai, Brian, and Clayton Thyne. "The effect of civil war on education, 1980—97." Journal of Peace Research 44, no. 3 (2007): 277-292.
- Levchenko, Andrei A. "Institutional quality and international trade." The Review of Economic Studies 74, no. 3 (2007): 791-819.
- Li, Daitian, Tony W. Tong, Yangao Xiao, and Feida Zhang. "Terrorism-induced uncertainty and firm R&D investment: A real options view." Journal of International Business Studies (2022): 1-13.
- Lindberg, Jonas, and Camilla Orjuela. "Corruption in the aftermath of war: an introduction." In Corruption in the Aftermath of War, pp. 1-14. Routledge, 2017.
- Markus, Keith E., Sahar Milani, and Rebecca Neumann. "The impact of patent protection and financial development on industrial R&D." Research Policy 48, no. 1 (2019): 355-370.
- Miguel, Edward, and Gerard Roland. "The long-run impact of bombing Vietnam." Journal of Development Economics 96, no. 1 (2011): 1-15.
- Milesi-Ferretti, Gian Maria, 2022, "The External Wealth of Nations Database," The Brookings Institution.
- Lane, Philip R., and Gian Maria Milesi-Ferretti. "The external wealth of nations revisited: international financial integration in the aftermath of the global financial crisis." IMF Economic Review 66 (2018): 189-222.
- Novta, Natalija, and Evgenia Pugacheva. "The macroeconomic costs of conflict." Journal of Macroeconomics 68 (2021): 103286.
- Nunn, Nathan. "Relationship-specificity, incomplete contracts, and the pattern of trade." The Quarterly Journal of Economics 122, no. 2 (2007): 569-600.
- OECD. STAN Bilateral Trade Database.
- Paunov, Caroline. "Corruption's asymmetric impacts on firm innovation." Journal of Development Economics 118 (2016): 216-231.
- Plümper, Thomas, and Eric Neumayer. "The unequal burden of war: The effect of armed conflict on the gender gap in life expectancy." International Organization 60, no. 3 (2006): 723-754.
- Raddatz, Claudio. "Liquidity needs and vulnerability to financial underdevelopment." Journal of Financial Economics 80, no. 3 (2006): 677-722.
- Rajan, Raghuram G., and Luigi Zingales. "Financial dependence and growth." The American Economic Review 88, no. 3 (1998): 559.
- Rodrik, Dani. "Where did all the growth go? External shocks, social conflict, and growth collapses." Journal of Economic Growth 4 (1999): 385-402.
- Rodrik, Dani. "Unconditional convergence in manufacturing." The Quarterly Journal of Economics 128, no. 1 (2013): 165-204.
- Romer, Paul M. "Endogenous technological change." Journal of Political Economy 98, no. 5, Part 2 (1990): 71-102.
- Sever, Can. "The myth of economic recovery: The case of crises in neighboring countries." Economics Letters 164 (2018): 65-69.
- Serneels, Pieter, and Marijke Verpoorten. "The impact of armed conflict on economic performance: Evidence from Rwanda." Journal of Conflict Resolution 59, no. 4 (2015): 555-592.
- Shemyakina, Olga. "The effect of armed conflict on accumulation of schooling: Results from Tajikistan." Journal of Development Economics 95, no. 2 (2011): 186-200.
- Teulings, Coen N., and Nikolay Zubanov. "Is economic recovery a myth? Robust estimation of impulse responses." Journal of Applied Econometrics 29, no. 3 (2014): 497-514.
- United Nations Industrial Development Organization (UNIDO) Database. Vienna.
- World Bank. World Bank Enterprise Surveys Database. Washington DC.
- World Bank. World Development indicators Database. Washington DC.
- World Bank. World Development Report 2011. “Conflict, Security, and Development”. Washington DC.
- Xu, Gang, and Go Yano. "How does anti-corruption affect corporate innovation? Evidence from recent anti-corruption efforts in China." Journal of Comparative Economics 45, no. 3 (2017): 498-519.

### Appendix — Macroeconomic Relevance of Conflicts (Figure A.1)
- Sample and data:
  - Global sample over the period of 1980-2018.
  - Real GDP per capita is from the World Bank’s World Development Indicators database.
  - The dummy variable indicating the years of domestic conflicts is from the Major Episodes of Political Violence (MEPV) database from the Center for Systemic Peace.
- Estimation approach:
  - Results are based on local projections (Jorda 2005).
  - Chart reports the coefficient estimates (reflecting the change in GDP per capita in percentage points) for 10 years with the 90 percent confidence intervals from local projections.
- Local projections specification (as presented):
  - 푙표푔(퐺퐷푃 푝푒푟 푐푎푝푖푡푎
    푐,푡+푝−1
    ) =훽
    0
    푝
     퐶표푛푓푙푖푐푡
    푡−1
    +∑훽
    푟
    푝
     퐶표푛푓푙푖푐푡
    푡+푟−1

    푝
    푟=1
    +휃
    푐
    푝
    +휃
    푡
    푝
    +휖
    푐,푡
    푝
  - 푐 and 푡 stand for country and year, respectively.
  - Country (휃
    푐
    푝
    ) and year (휃
    푡
    푝
    ) fixed effects are included.
  - Standard errors are clustered at the country-level.
  - The regression is run for 10 years, i.e., 푝=1,....,10, controlling for the forward values of conflicts to eliminate potential bias (as proposed by Teulings and Zubanov 2014).
  - Figure A.1 reports the coefficient estimates of 훽
    0
    푝
    for the 10-year period in the aftermath of conflicts (i.e., 푝=1,....,10).

### Summary Statistics (Table A.1)
- Industry characteristics:
  - R&D intensity baseline: Mean 0.971, Median 0.422, 25th ptile 0.266, 75th ptile 0.926, Std. dev. 1.335
  - R&D intensity alternative I: Mean 0.0371, Median 0.0122, 25th ptile 0.0069, 75th ptile 0.0232, Std. dev. 0.0638
  - R&D intensity alternative II: Mean 0.016, Median 0.005, 25th ptile 0.004, 75th ptile 0.028, Std. dev. 0.020
  - Patent intensity: Mean 4.8, Median 0.6, 25th ptile 0.2, 75th ptile 6.4, Std. dev. 7.6
  - Quality certificate intensity: Mean 28.7, Median 28.6, 25th ptile 16.2, 75th ptile 41.1, Std. dev. 13.5
  - Physical capital intensity: Mean 1.204, Median 1.117, 25th ptile 0.911, 75th ptile 1.466, Std. dev. 0.520
  - Product complexity (inverse Herfindahl): Mean -0.083, Median -0.077, 25th ptile -0.106, 75th ptile -0.060, Std. dev. 0.030
  - Dependence on external finance: Mean -0.096, Median -0.262, 25th ptile -0.385, 75th ptile -0.140, Std. dev. 1.164
  - Liquidity needs: Mean 0.166, Median 0.177, 25th ptile 0.126, 75th ptile 0.203, Std. dev. 0.050
  - Asset tangibility: Mean 0.280, Median 0.259, 25th ptile 0.204, 75th ptile 0.363, Std. dev. 0.126
  - Import share: Mean 0.045, Median 0.035, 25th ptile 0.027, 75th ptile 0.062, Std. dev. 0.031
  - Export share: Mean 0.046, Median 0.033, 25th ptile 0.020, 75th ptile 0.051, Std. dev. 0.037
- Industry-level variables:
  - Labor productivity growth (%): Mean 4.10, Median 4.50, 25th ptile -7.09, 75th ptile 15.79, Std. dev. 27.72
  - Value added share (%): Mean 5.48, Median 3.43, 25th ptile 1.47, 75th ptile 6.63, Std. dev. 7.10
- Notes: The variables are explained in detail in Section 2.

### Correlations between measures of technological intensity (Table A.2)
- Correlation matrix (selected entries):
  - RDI baseline with RDI alternative I: 0.98*** 
  - RDI baseline with RDI alternative II: 0.75***
  - RDI baseline with Patent intensity: 0.65***
  - RDI baseline with QC intensity: 0.45**
  - RDI alternative I with RDI alternative II: 0.69***
  - RDI alternative I with Patent intensity: 0.56***
  - RDI alternative I with QC intensity: 0.40*
  - RDI alternative II with Patent intensity: 0.91***
  - RDI alternative II with QC intensity: 0.74***
  - Patent intensity with QC intensity: 0.62***
- Notes: *** p<0.01, ** p<0.05, * p<0.1. Variables are explained in detail in Section 2.

### Countries in the Sample
- The list of countries (restricted by the availability of data) is:
  Albania, Algeria, Argentina, Armenia, Australia, Austria, Azerbaijan, Bahrain, Bangladesh, Belarus, Belgium, Bolivia, Bosnia and Herzegovina, Botswana, Brazil, Bulgaria, Burundi, Cameroon, Canada, Central African Republic, Chile, China, Colombia, Costa Rica, Croatia, Cuba, Cyprus, Czechia, Denmark, Ecuador, Egypt, El Salvador, Eritrea, Estonia, Eswatini, Ethiopia, Fiji, Finland, France, Georgia, Germany, Ghana, Greece, Guatemala, Honduras, Hungary, India, Indonesia, Iran, Iraq, Ireland, Israel, Italy, Jamaica, Japan, Jordan, Kazakhstan, Kenya, Korea, Kuwait, Kyrgyzstan, Lao, Latvia, Lithuania, Luxembourg, Madagascar, Malawi, Malaysia, Mauritius, Mexico, Moldova, Mongolia, Montenegro, Morocco, Netherlands, New Zealand, Niger, North Macedonia, Norway, Oman, Pakistan, Panama, Papua New Guinea, Paraguay, Peru, Philippines, Poland, Portugal, Qatar, Romania, Russia, Saudi Arabia, Senegal, Singapore, Slovakia, Slovenia, South Africa, Spain, Sri Lanka, Sweden, Switzerland, Syria, Tanzania, Thailand, Trinidad and Tobago, Tunisia, Türkiye, UK, Uruguay, USA, Venezuela, Vietnam, Yemen, Zambia.

### Accounting for Industry Size (Table A.3)
- Selected coefficient estimates and statistics (standard errors in parentheses):
  - 푇푒푐ℎ 푖푛푡푒푛푠푖푡푦
    푗
    ×퐶표푛푓푙푖푐푡
    푐,푡+2 : 0.755 (0.585)
  - 푇푒푐ℎ 푖푛푡푒푛푠푖푡푦
    푗
    ×퐶표푛푓푙푖푐푡
    푐,푡+1 : 0.600 (0.808)
  - 푇푒푐ℎ 푖푛푡푒푛푠푖푡푦
    푗
    ×퐶표푛푓푙푖푐푡
    푐,푡 : -2.238*** (0.777)
  - 푇푒푐ℎ 푖푛푡푒푛푠푖푡푦
    푗
    ×퐶표푛푓푙푖푐푡
    푐,푡−1 : 0.748 (0.918)
  - 푇푒푐ℎ 푖푛푡푒푛푠푖푡푦
    푗
    ×퐶표푛푓푙푖푐푡
    푐,푡−2 : 0.611 (0.661)
  - 푉퐴 푠ℎ푎푟푒
    푐,푗,푡−1 ×퐶표푛푓푙푖푐푡
    푐,푡+2 : 0.001 (0.130)
  - 푉퐴 푠ℎ푎푟푒
    푐,푗,푡−1 ×퐶표푛푓푙푖푐푡
    푐,푡+1 : 0.111 (0.130)
  - 푉퐴 푠ℎ푎푟푒
    푐,푗,푡−1 ×퐶표푛푓푙푖푐푡
    푐,푡 : -0.079 (0.144)
  - 푉퐴 푠ℎ푎푟푒
    푐,푗,푡−1 ×퐶표푛푓푙푖푐푡
    푐,푡−1 : 0.056 (0.105)
  - 푉퐴 푠ℎ푎푟푒
    푐,푗,푡−1 ×퐶표푛푓푙푖푐푡
    푐,푡−2 : -0.162 (0.120)
  - 푉퐴 푠ℎ푎푟푒
    푐,푗,푡−1 : -1.487*** (0.089)
- Fixed effects:
  - Country-year F.E.: Yes
  - Country-industry F.E.: Yes
  - Industry-year F.E.: Yes
- Observations: 49,518
- R-squared: 0.358
- Notes: Results are based on equation 2. Standard errors in parentheses are clustered at the country-industry level. *** p<0.01, ** p<0.05, * p<0.1.

*Content derived from the source PDF: wpiea2024098-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024098-print-pdf.pdf_