## section 5 concludes.

## Source details

**Canonical URL:** [section 5 concludes.](https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026122-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2026/english/wpiea2026122-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2026/english/wpiea2026122-source-pdf.pdf.json)

---

### Methodology
- Identification strategy: staggered difference-in-differences design. Firms enter treatment in the year they first receive state aid and remain treated thereafter; never-received firms serve as controls. Unbalanced panel covering 2016 to 2023.
- Baseline estimator: Callaway and SantAnna (2021) cohort-specific event-study aggregation:
  - ATT(c, k) = E[y_{i,t+k}(1) − y_{i,t+k}(0) | i ∈ c]
  - ATT(k) = ∑_{c∈C} ATT(c, k) · P(G_i = c | i ∈ C_k)
  - Allows heterogeneous treatment effects across cohorts and dynamic effects by event time.
- Heterogeneity analysis: separate regressions for cohorts of young firms and small firms.
- Pre-treatment comparability: cohort-by-cohort propensity score matching implemented (details in appendix).
- Comparative specification: conventional two-way fixed-effects:
  - log(y_it) = β TreatedPost_it + β_y TreatedPost_it × Young_i + α_i + γ_st + ε_it
  - Includes firm fixed effects α_i and sector-year fixed effects γ_st; noted bias when treatment effects heterogeneous across cohorts motivates Callaway and SantAnna benchmark.
- Event-study specification for parallel trends testing:
  - log(y_it) = ∑_k β_k · 1{EventTime_it = k} + ∑_k β_{k,y} · 1{EventTime_it = k} × Young_i + α_i + γ_st + ε_it
  - EventTime_it = year_t − TreatmentYear_i, with EventTime = −1 omitted as baseline.

### Data
- Main sources:
  - Firm-level balance-sheet data from Orbis (Bureau van Dijk), years 2016 to 2023. Nominal values deflated using Federal Reserve Economic Data GDP deflator.
  - State Aid Transparency database (France): around 11,000 unique firms received state aid between 2016 and 2023; includes granting date, amount, recipient national identifier, objective, and instrument type (guarantees, loans, direct grants, interest rate subsidies, tax advantages).
  - EUTL-Orbis matched dataset for firm-level verified emissions and allocation of free allowances under the EU ETS.
- Sample construction and exclusions:
  - Exclude awards under the objective of remedy for serious disturbance of the economy.
  - Exclude firms receiving aid in multiple years to simplify pre/post categorization.
  - Merged sample: more than 25 percent firms are young at aid receipt; approximately 85 percent are small.
  - Among firms in the merged sample with aid amounts available, young firms account for an average of roughly 50 percent of total aid awarded.
- Reporting thresholds (data caveats): reporting threshold was 500,000 euros until 2023 after which it was revised to 100,000 euros.
- Data limitations:
  - State aid information only from 2016 onward — some control firms may have received prior aid.
  - Transparency data is self-reported and only includes aid above thresholds — potential under-reporting of smaller recipients could bias estimates toward underestimating treatment effects.
  - Orbis limited coverage of small and young firms, though share of small firms in merged sample comparable to Transparency Database.

### Main empirical findings
- Overall average effect on sales:
  - Average effect on sales is small and statistically insignificant: 2.4 percent.
- Strong heterogeneity by firm age:
  - Young firms (aged below 10 years at treatment) experience, on average, a large increase of 71 percent in sales, with the effect emerging immediately after treatment and remaining economically and statistically significant throughout the sample period.
  - Older firms show small, negative, and statistically insignificant responses for sales, productivity, and employment.
- Heterogeneity by firm size:
  - Targeting by firm size appears less effective:
    - Small firms experience modest improvements; sales increase by roughly 12 percent above the overall estimate, but effects are smaller and often statistically insignificant relative to young firms.
    - Large firms’ sales decline on average, though estimates are statistically insignificant.
- Private financing / crowding-in:
  - Young firms increase their external debt in years following state aid, suggesting potential crowding-in of private financing; older firms show little change in borrowing behavior.
  - Note: part of the rise in external debt may reflect expanded pandemic support measures not fully captured in the data.
- Sectoral targeting and aid objectives:
  - R&D aid: strongest response in high-technology sectors, particularly for young firms; little measurable effect outside high-tech sectors.
  - Energy aid: strongest impact in manufacturing, with positive effects across firms but more pronounced for young beneficiaries.
- Emissions and the green transition (associations, not causal):
  - Sample: EU ETS-compliant manufacturing firms in France receiving energy and environmental aid; emission intensity defined as verified ETS allowances (tons of CO2) per thousand USD of sales.
  - Regression: Emission Intensity_i,t = β_0 + β_1 price_C,t + β_2 aid_i,t + β_3 price_C,t × aid_i,t + α_i (with firm fixed effects).
  - Estimates:
    - A one dollar increase in carbon prices tends to lower emission intensity on average by 0.01 tons of CO2 per $1,000 of output.
    - The reduction is twice as large for firms receiving energy and environmental aid: around 0.02 tons of CO2 per $1,000 of output.
  - Interpretation: findings point to a transition channel where state aid helps firms reduce effective marginal abatement costs; caution urged because the carbon emissions sample is skewed toward large manufacturing firms and sample sizes are small.

### Policy implications and recommendations
- Prioritize age-based targeting:
  - State aid is most effective when directed to young firms (below 10 years), which exhibit large and persistent gains in sales, productivity, and employment and show crowding-in of private debt.
- Reassess size-based targeting:
  - Size-based targeting (small vs large firms) appears less effective; size-based policies risk distorting firm sizes and producing inefficient allocation.
- Match aid objectives to sectoral conditions:
  - R&D support is most effective for young firms in high-technology sectors; energy aid is most effective in manufacturing, especially for young beneficiaries.
- Design aid to support the green transition:
  - Energy and environmental aid can amplify the emissions-reducing effect of carbon pricing (suggestive evidence of a larger emission-intensity response among aided firms); consider aid that lowers upfront costs of adopting lower-emission technologies rather than simply compensating carbon costs.
- Balance firm-level gains with market effects:
  - Firm-level benefits (especially for young firms) should be weighed against possible negative spillovers on competitors and broader market distortions.

### Appendix — key statistics and robustness checks
- Summary statistics (Panel A: Full sample):
  - Log sales: Number of observations 6,249,508; Mean 12.899; Standard deviation 1.975
  - Log employment: Number of observations 525,155; Mean 3.527; Standard deviation 1.097
  - Log productivity: Number of observations 1,621,486; Mean 12.083; Standard deviation 0.955
  - Log debt: Number of observations 4,999,414; Mean 10.926; Standard deviation 2.361
- Regression sample (Panel B: after propensity score matching):
  - Log sales: Number of observations 14,290; Mean 16.672; Standard deviation 2.377
  - Log employment: Number of observations 8,206; Mean 4.905; Standard deviation 1.837
  - Log productivity: Number of observations 9,239; Mean 12.506; Standard deviation 1.118
  - Log debt: Number of observations 14,033; Mean 14.870; Standard deviation 2.500
- Emissions sample (Panel C):
  - Log sales: Number of observations 3,865; Mean 17.893; Standard deviation 1.978
  - Emission intensity: Number of observations 3,224; Mean 0.001; Standard deviation 0.181
- Sample composition by treatment status (Number of unique firms):
  - Full sample: Control group 1,632,255; Treatment group 3,871; Total 1,636,126
  - PSM sample: Control group 1,504; Treatment group 353; Total 1,857
  - Emission sample: Control group 383; Treatment group 624; Total 45
- Coverage by size (Transparency database totals):
  - Total: Large firms 1,481; SME 9,875; Total 11,356; Share of SME 87.0
- Coverage by size (Transparency-Orbis merged database totals):
  - Total: Large firms 557; SME 3,314; Total 3,871; Share of SME 85.6
- Two-way fixed effect results (selected):
  - Specification 1: Baseline — Treated: Sales 0.272***; Productivity 0.077***; Employment 0.018; Debt 0.141***; Observations: 1,840,102; 404,664; 200,331; 149,1536
  - Specification 2: Age — Treated×Young: Sales 0.526***; Productivity 0.080; Employment 0.181***; Debt 0.214***; Treated: Sales 0.137***; Productivity 0.066***; Employment -0.003; Debt 0.086***; Observations: 1,840,102; 404,664; 200,331; 149,1536
  - Specification 3: Size — Treated×Small: Sales -0.106***; Productivity 0.147***; Employment -0.056; Debt 0.071; Treated: Sales 0.297***; Productivity 0.020; Employment 0.037***; Debt 0.124***; Observations: 1,840,102; 404,664; 200,331; 149,1536
- Robustness controlling for aid amount quartiles:
  - Treated: Sales 0.118***; Productivity 0.007; Employment -0.035; Debt -0.004
  - Treated×Young: Sales 0.495***; Productivity 0.130; Employment 0.188***; Debt 0.211***
  - Observations: 11,822; 5,409; 5,413; 1,657
- Sector and age interactions (selected from Table 6):
  - High-tech: Young×R&D aid: Productivity 0.665**; Employment 0.287**
  - Manufacturing: Young×R&D aid: Employment 0.589***; Young×Energy aid: Employment 0.160***
- Energy aid effect on emission intensity (Table 7, Observations 3,100):
  - L.Carbon price: -0.0111*
  - Recipient: -0.1896
  - L.Carbon price × recipient: -0.0107***
  - Firm FE: Yes; Year FE: Yes; Clustered SE: Nace 4
- Methodological and implementation notes:
  - Aid variable treated as binary in main analysis.
  - Cohort-by-cohort PSM with nearest-neighbor matching (five neighbors) and analytic weights applied in event-study re-estimation.
  - Programming: csdid (Stata) package on Stata version 18.5 and 19.5.

*Source: section 5 concludes. wpiea2026122-source-pdf - section 5 concludes.*

### section 5 concludes.

### section 5 concludes.

### Methodology
- Identification strategy: staggered difference-in-differences design. Firms enter treatment in the year they first receive state aid and remain treated thereafter; never-received firms serve as controls. Unbalanced panel covering 2016 to 2023.
- Baseline estimator: Callaway and SantAnna (2021) cohort-specific event-study aggregation:
  - ATT(c, k) = E[y_{i,t+k}(1) − y_{i,t+k}(0) | i ∈ c]
  - ATT(k) = ∑_{c∈C} ATT(c, k) · P(G_i = c | i ∈ C_k)
  - Allows heterogeneous treatment effects across cohorts and dynamic effects by event time.
- Heterogeneity analysis: separate regressions for cohorts of young firms and small firms.
- Pre-treatment comparability: cohort-by-cohort propensity score matching implemented (details in appendix).
- Comparative specification: conventional two-way fixed-effects:
  - log(y_it) = β TreatedPost_it + β_y TreatedPost_it × Young_i + α_i + γ_st + ε_it
  - Includes firm fixed effects α_i and sector-year fixed effects γ_st; noted bias when treatment effects heterogeneous across cohorts motivates Callaway and SantAnna benchmark.
- Event-study specification for parallel trends testing:
  - log(y_it) = ∑_k β_k · 1{EventTime_it = k} + ∑_k β_{k,y} · 1{EventTime_it = k} × Young_i + α_i + γ_st + ε_it
  - EventTime_it = year_t − TreatmentYear_i, with EventTime = −1 omitted as baseline.

### Data
- Main sources:
  - Firm-level balance-sheet data from Orbis (Bureau van Dijk), years 2016 to 2023. Nominal values deflated using Federal Reserve Economic Data GDP deflator.
  - State Aid Transparency database (France): around 11,000 unique firms received state aid between 2016 and 2023; includes granting date, amount, recipient national identifier, objective, and instrument type (guarantees, loans, direct grants, interest rate subsidies, tax advantages).
  - EUTL-Orbis matched dataset for firm-level verified emissions and allocation of free allowances under the EU ETS.
- Sample construction and exclusions:
  - Exclude awards under the objective of remedy for serious disturbance of the economy.
  - Exclude firms receiving aid in multiple years to simplify pre/post categorization.
  - Merged sample: more than 25 percent firms are young at aid receipt; approximately 85 percent are small.
  - Among firms in the merged sample with aid amounts available, young firms account for an average of roughly 50 percent of total aid awarded.
- Reporting thresholds (data caveats): reporting threshold was 500,000 euros until 2023 after which it was revised to 100,000 euros.
- Data limitations:
  - State aid information only from 2016 onward — some control firms may have received prior aid.
  - Transparency data is self-reported and only includes aid above thresholds — potential under-reporting of smaller recipients could bias estimates toward underestimating treatment effects.
  - Orbis limited coverage of small and young firms, though share of small firms in merged sample comparable to Transparency Database.

### Main empirical findings
- Overall average effect on sales:
  - Average effect on sales is small and statistically insignificant: 2.4 percent.
- Strong heterogeneity by firm age:
  - Young firms (aged below 10 years at treatment) experience, on average, a large increase of 71 percent in sales, with the effect emerging immediately after treatment and remaining economically and statistically significant throughout the sample period.
  - Older firms show small, negative, and statistically insignificant responses for sales, productivity, and employment.
- Heterogeneity by firm size:
  - Targeting by firm size appears less effective:
    - Small firms experience modest improvements; sales increase by roughly 12 percent above the overall estimate, but effects are smaller and often statistically insignificant relative to young firms.
    - Large firms’ sales decline on average, though estimates are statistically insignificant.
- Private financing / crowding-in:
  - Young firms increase their external debt in years following state aid, suggesting potential crowding-in of private financing; older firms show little change in borrowing behavior.
  - Note: part of the rise in external debt may reflect expanded pandemic support measures not fully captured in the data.
- Sectoral targeting and aid objectives:
  - R&D aid: strongest response in high-technology sectors, particularly for young firms; little measurable effect outside high-tech sectors.
  - Energy aid: strongest impact in manufacturing, with positive effects across firms but more pronounced for young beneficiaries.
- Emissions and the green transition (associations, not causal):
  - Sample: EU ETS-compliant manufacturing firms in France receiving energy and environmental aid; emission intensity defined as verified ETS allowances (tons of CO2) per thousand USD of sales.
  - Regression: Emission Intensity_i,t = β_0 + β_1 price_C,t + β_2 aid_i,t + β_3 price_C,t × aid_i,t + α_i (with firm fixed effects).
  - Estimates:
    - A one dollar increase in carbon prices tends to lower emission intensity on average by 0.01 tons of CO2 per $1,000 of output.
    - The reduction is twice as large for firms receiving energy and environmental aid: around 0.02 tons of CO2 per $1,000 of output.
  - Interpretation: findings point to a transition channel where state aid helps firms reduce effective marginal abatement costs; caution urged because the carbon emissions sample is skewed toward large manufacturing firms and sample sizes are small.

### Policy implications and recommendations
- Prioritize age-based targeting:
  - State aid is most effective when directed to young firms (below 10 years), which exhibit large and persistent gains in sales, productivity, and employment and show crowding-in of private debt.
- Reassess size-based targeting:
  - Size-based targeting (small vs large firms) appears less effective; size-based policies risk distorting firm sizes and producing inefficient allocation.
- Match aid objectives to sectoral conditions:
  - R&D support is most effective for young firms in high-technology sectors; energy aid is most effective in manufacturing, especially for young beneficiaries.
- Design aid to support the green transition:
  - Energy and environmental aid can amplify the emissions-reducing effect of carbon pricing (suggestive evidence of a larger emission-intensity response among aided firms); consider aid that lowers upfront costs of adopting lower-emission technologies rather than simply compensating carbon costs.
- Balance firm-level gains with market effects:
  - Firm-level benefits (especially for young firms) should be weighed against possible negative spillovers on competitors and broader market distortions.

*Source: section 5 concludes. wpiea2026122-source-pdf - section 5 concludes.*

### References

### References and Appendix (wpiea2026122-source-pdf)

### Key bibliographic references
- Acemoglu, Daron, Ufuk Akcigit, Douglas Hanley, and William Kerr (2016) “Transition to clean technology,” Journal of political economy, 124 (1), 52–104.
- Aghion, Philippe, Antonin Bergeaud, and John Van Reenen (2023) “The impact of regulation on innovation,” American Economic Review, 113 (11), 2894–2936.
- Bajgar, Matej, Giuseppe Berlingieri, Sara Calligaris, Chiara Criscuolo, and Jonathan Timmis (2020) “Coverage and representativeness of Orbis data,” OECD Science, Technology and Industry Working Papers.
- Baquie, Sandra, Yueling Huang, Ms Florence Jaumotte, Jaden Kim, Rafael Machado Parente, and Samuel Pienknagura (2025) Industrial policies: Handle with care: International Monetary Fund.
- Brandão-Marques, Mr Luis and Hasan H Toprak (2024) A Bitter Aftertaste: How State Aid Affects Recipient Firms and Their Competitors in Europe: International Monetary Fund.
- Cabral, Luís M B and Jose Mata (2003) “On the evolution of the firm size distribution: Facts and theory,” American economic review, 93 (4), 1075–1090.
- Callaway, Brantly and Pedro HC SantAnna (2021) “Difference-in-differences with multiple time periods,” Journal of econometrics, 225 (2), 200–230.
- Cameron, Alienor and Maria Garrone (2024) “Carbon intensity and corporate performance: A micro-level study of EU ETS industrial firms.”
- Colmer, Jonathan, Ralf Martin, Mirabelle Muûls, and Ulrich J Wagner (2025) “Does pricing carbon mitigate climate change? firm-level evidence from the european union emissions trading system,” Review of Economic Studies, 92 (3), 1625–1660.
- Criscuolo, Chiara, Ralf Martin, Henry G Overman, and John Van Reenen (2019) “Some causal effects of an industrial policy,” American Economic Review, 109 (1), 48–85.
- D’Arcangelo, Filippo, Levin Ilai, Alessia Pagani, Mauro Pisu, and Åsa Johansson (2022) “A framework to decarbonise the economy,” OECD Economic Policy Papers, No.31, 42.
- Foda, Karim, Yu Shi, and Maryam Vaziri (2025) “Financial constraints, firm dynamics and productivity,” Applied Economics Letters, 32 (10), 1439–1447.
- Garicano, Luis, Claire Lelarge, and John Van Reenen (2016) “Firm size distortions and the productivity distribution: Evidence from France,” American Economic Review, 106 (11), 3439–3479.
- Goodman-Bacon, Andrew (2021) “Difference-in-differences with variation in treatment timing,” Journal of econometrics, 225 (2), 254–277.
- Hadlock, Charles J and Joshua R Pierce (2010) “New evidence on measuring financial constraints: Moving beyond the KZ index,” The review of financial studies, 23 (5), 1909–1940.
- Hodge, Andrew, Mr Roberto Piazza, Fuad Hasanov, Xun Li, Maryam Vaziri, Mr Atticus Weller, and Yu Ching Wong (2024) Industrial policy in europe: A single market perspective: International Monetary Fund.
- IMF (2024) “INDUSTRIAL POLICY COVERAGE IN IMF SURVEILLANCE.”
- IMF (2025) Industrial Policy: Managing Trade-offs to promote growth and resilience: IMF.
- Juhász, Réka and Nathan Lane (2024) “The political economy of industrial policy,” Journal of Economic Perspectives, 38 (4), 27–54.
- Juhász, Réka, Nathan Lane, and Dani Rodrik (2024) “The new economics of industrial policy,” Annual Review of Economics, 16 (1), 213–242.
- Lane, Nathan (2025) “Manufacturing revolutions: Industrial policy and industrialization in South Korea,” The Quarterly Journal of Economics, 140 (3), 1683–1741.
- Malessan, Anna (2025) “The Impact of Subsidizing early Young Innovative Companies on their Access to Capital Market,” Technical report, TEPP.
- Rodrik, Dani (2004) “Industrial policy for the twenty-first century.”

### Appendix overview
- The appendix is organized as follows:
  - Description of the data and summary statistics.
  - Documentation of sample construction and cleaning steps.
  - Supplementary results not reported in the main text.

### A.1 Data and coverage — summary statistics and sample composition
- Table 1: Summary statistics of outcome variables
  - Panel A: Full sample
    - Log sales: Number of observations 6,249,508; Mean 12.899; Standard deviation 1.975
    - Log employment: Number of observations 525,155; Mean 3.527; Standard deviation 1.097
    - Log productivity: Number of observations 1,621,486; Mean 12.083; Standard deviation 0.955
    - Log debt: Number of observations 4,999,414; Mean 10.926; Standard deviation 2.361
  - Panel B: Regression sample (after propensity score matching)
    - Log sales: Number of observations 14,290; Mean 16.672; Standard deviation 2.377
    - Log employment: Number of observations 8,206; Mean 4.905; Standard deviation 1.837
    - Log productivity: Number of observations 9,239; Mean 12.506; Standard deviation 1.118
    - Log debt: Number of observations 14,033; Mean 14.870; Standard deviation 2.500
  - Panel C: Emissions sample
    - Log sales: Number of observations 3,865; Mean 17.893; Standard deviation 1.978
    - Emission intensity: Number of observations 3,224; Mean 0.001; Standard deviation 0.181

- Table 2: Sample composition by treatment status (Number of unique firms)
  - Full sample: Control group 1,632,255; Treatment group 3,871; Total 1,636,126
  - PSM sample: Control group 1,504; Treatment group 353; Total 1,857
  - Emission sample: Control group 383; Treatment group 624; Total 45

- Coverage by size (Transparency database and Transparency-Orbis merged database)
  - Transparency database (selected rows and totals)
    - 2016: Large firms 83; SME 356; Total 439; Share of SME 81.1
    - 2017: Large firms 179; SME 882; Total 1,061; Share of SME 83.1
    - 2018: Large firms 146; SME 927; Total 1,073; Share of SME 86.4
    - 2019: Large firms 105; SME 808; Total 913; Share of SME 88.5
    - 2020: Large firms 126; SME 1,001; Total 1,127; Share of SME 88.8
    - 2021: Large firms 373; SME 2,696; Total 3,069; Share of SME 87.8
    - 2022: Large firms 305; SME 2,426; Total 2,731; Share of SME 88.8
    - 2023: Large firms 164; SME 779; Total 943; Share of SME 82.6
    - Total: Large firms 1,481; SME 9,875; Total 11,356; Share of SME 87.0
  - Transparency-Orbis merged database (selected rows and totals)
    - 2016: Large firms 27; SME 170; Total 197; Share of SME 86.3
    - 2017: Large firms 76; SME 432; Total 508; Share of SME 85.0
    - 2018: Large firms 62; SME 388; Total 450; Share of SME 86.2
    - 2019: Large firms 23; SME 350; Total 373; Share of SME 93.8
    - 2020: Large firms 39; SME 372; Total 411; Share of SME 90.5
    - 2021: Large firms 141; SME 886; Total 1,027; Share of SME 86.3
    - 2022: Large firms 148; SME 588; Total 736; Share of SME 79.9
    - 2023: Large firms 41; SME 128; Total 169; Share of SME 75.7
    - Total: Large firms 557; SME 3,314; Total 3,871; Share of SME 85.6

- Note: SMEs account for 85.6 percent of firms in the merged sample versus 87 percent in the full database.

### A.2 Methodological details and sample construction
- Sample exclusions:
  - Exclude firms in NACE sections P to U (education, health, arts, and other non-market services).
- Sector classification:
  - High-tech sector: Information and Communication (NACE Rev. 2, Section J) and Professional, Scientific and Technical Activities (Section M).
  - Manufacturing sector: Manufacturing (Section C) and Electricity, Gas, Steam and Air Conditioning Supply (Section D).
- Aid variable:
  - State aid treated as a binary variable due to aid amounts sometimes reported as ranges.
- Aid classification:
  - Distinguish energy, energy and environmental, and research and development aid via keywords extracted from objectives and titles (full lists in Figures 6, 7, and 8).
- Outcome variables:
  - Log transforms applied above minimal positive thresholds (y_it > 1).
- Propensity score matching (PSM):
  - Cohort-by-cohort PSM.
  - Baseline year b = g − 1; firm treated in cohort b if aid receipt occurs in b + 1.
  - Control group includes cohort-treated firms and firms never treated.
  - Propensity score uses pre-treatment firm characteristics: sales, sales growth over three years, total assets (all in logarithm and lagged), firm age and number of employees in logarithm, and sector fixed effects.
  - Nearest-neighbor matching (five neighbors) on the region of common support.
  - Firm-level weight constructed: 1 for treated firms; match weight for controls.
  - Re-estimate event-study on matched sample with analytic weights, absorbing firm fixed effects (α_i) and sector-year fixed effects (γ_st).
- Programming:
  - Used csdid (Stata) package on Stata version 18.5 and 19.5.

### A.3 Other specifications — additional estimation results
- Table 4: Two-way fixed effect outcomes (Note: Standard errors robust. *p<0.10, **p<0.05, ***p<0.01. Young firms are less than 10 years old in the year they receive aid. Small firms have fewer than 50 employees in the year they receive aid. All outcomes are in logs.)
  - Specification 1: Baseline
    - Treated: Sales 0.272***; Productivity 0.077***; Employment 0.018; Debt 0.141***
    - Observations: 1,840,102; 404,664; 200,331; 149,1536
  - Specification 2: Age
    - Treated×Young: Sales 0.526***; Productivity 0.080; Employment 0.181***; Debt 0.214***
    - Treated: Sales 0.137***; Productivity 0.066***; Employment -0.003; Debt 0.086***
    - Observations: 1,840,102; 404,664; 200,331; 149,1536
  - Specification 3: Size
    - Treated×Small: Sales -0.106***; Productivity 0.147***; Employment -0.056; Debt 0.071
    - Treated: Sales 0.297***; Productivity 0.020; Employment 0.037***; Debt 0.124***
    - Observations: 1,840,102; 404,664; 200,331; 149,1536
  - Firm FE: Yes for all; Sector×Year FE: Yes for all.

- Table 5: Two-way fixed effect outcomes controlling for received amount in quartiles when available (Less than 6 percent of sample reported aid amount in a range and are eliminated from this regression.)
  - Treated: Sales 0.118***; Productivity 0.007; Employment -0.035; Debt -0.004
  - Treated×Young: Sales 0.495***; Productivity 0.130; Employment 0.188***; Debt 0.211***
  - Firm FE: Yes; Sector×Year FE: Yes; Aid amount FE: Yes.
  - Observations: 11,822; 5,409; 5,413; 1,657

- Table 6: Effects of R&D and energy aid on employment and productivity by sector and firm age
  - High-tech (Observations 14,171; 1,061)
    - R&D aid: Productivity 0.045; Employment 0.131**
    - Young×R&D aid: Productivity 0.665**; Employment 0.287**
    - Energy aid: Productivity -0.283; Employment 0.239**
    - Young×Energy aid: Productivity -0.305; Employment -0.160*
  - Manufacturing (Observations 21,301; 8,38)
    - R&D aid: Productivity 0.179**; Employment -0.112
    - Young×R&D aid: Productivity -0.451*; Employment 0.589***
    - Energy aid: Productivity -0.151; Employment -0.033
    - Young×Energy aid: Productivity 0.247; Employment 0.160***

- Table 7: Energy Aid Effect on Emission Intensity
  - Dependent variable: Emission Intensity (Observations 3,100)
  - L.Carbon price: -0.0111*
  - Recipient: -0.1896
  - L.Carbon price × recipient: -0.0107***
  - Firm FE: Yes; Year FE: Yes; Clustered SE: Nace 4

- Figure 9: Event studies
  - Note: Estimates report average treatment effects using the full sample. Whiskers indicate 95 percent confidence intervals.

*Targeting State Aid: Firm-level Evidence from France — Working Paper No. WP/2026/122*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2026/english/wpiea2026122-source-pdf.pdf_