## wpiea2024250-print-pdf

## Source details

**Canonical URL:** [wpiea2024250-print-pdf](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024250-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024250-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024250-print-pdf.pdf.json)

---

### Executive summary — core findings
- State aid can temporarily boost firm-level employment and revenue of recipient firms but has no discernible impact on firm investment or productivity.
- Measured effects on recipient firms:
  - employment increases by 0.3 percent in the following year.
  - revenues increase by 0.6 percent in the following year.
- Adverse spillovers to competitor (nonrecipient) firms:
  - decreases in employment, revenues, and productivity in the same sector.
  - no detectable effect on investment for competitors.
- Magnitude and persistence of spillovers:
  - Adverse spillovers are at least half as large as the own-firm effects after one year.
  - Spillovers are more long-lasting and overturn the initial positive own-firm effects after two years.
- Heterogeneity of own-firm effects:
  - Stronger effects for firms that are smaller, have less liquidity, lower debt-servicing capacity, and are younger.
  - No differential response for firms with higher R&D intensity.
- Interpretation:
  - State aid may overcome firm-level financial constraints but does not appear to enhance firms’ ability to innovate or raise productivity.
  - Large-scale state aid programs would unlikely raise aggregate productivity in Europe given the negative spillovers.

### Policy implications and recommendations
- National-level state aid risks distorting competition within the EU single market and can create an uneven playing field that favors domestic firms over competitors from other member states.
- Net aggregate losses for the EU are possible because negative spillovers from state aid often surpass short-term benefits to recipients.
- Recommendation: when state aid is justified by well-identified externalities and less distortionary tools are absent, coordination at the EU level is preferable.
  - Pooling resources and competitively allocating aid across the Union could preserve market competition, encourage firm entry, and ensure a more efficient distribution of funds.
  - An EU-level approach can better capture gains across value chains spanning multiple member states, reduce capture by entrenched interests, and address disparities in fiscal capacity.
- An EU-coordinated industrial policy could help firms achieve economies of scale that would be difficult under fragmented national interventions, particularly amid increasing geoeconomic fragmentation.

### Research design and identification
- Scope and sample:
  - State aid awarded between 2016 and 2023 to listed nonfinancial firms in Belgium, France, Germany, Italy, the Netherlands, Spain, and the United Kingdom (until 2020).
  - State aid data from the European Commission’s State Aid Transparency Database; firm accounting and employment data from Orbis; excess equity returns (daily) from Bloomberg.
- Identification strategy:
  - High-frequency identification approach using excess equity returns of recipient firms on announcement dates to capture the nonanticipated component of state aid and address endogeneity under market efficiency assumptions.
  - Key assumptions: financial markets share government information; systematic policy components are anticipated and priced before the announcement; post-announcement equity return surprises identify exogenous variation.
- Measurement:
  - Firm-specific state aid surprise = daily return on the stock of each firm in excess of the average daily market return on the date of granting (date of granting = day when legal right to receive the aid is conferred).
  - State aid shock expressed in percent; symmetric around zero with most surprises small; reported not correlated with observable firm characteristics.
- State aid spillover (SAS):
  - For each firm: the average of state aid shocks (weighted by firm assets) in all other countries in a given year and in the same industry (NACE level 4).

### Econometric specifications and outcomes
- Benchmark specifications:
  - Own-effects (equation (1)): y_{i,c,t+1} − y_{i,c,t} = α^y + β^y_o × SA_{i,c,t} + γ^y × X_{i,c,t} + ε_{i,c,t}
  - Spillover-effects (equation (2)): y_{i,c,t+1} − y_{i,c,t} = α^y + β^y_s × SAS_{i,c,t} + γ^y × X_{i,c,t} + ε_{i,c,t}
  - Combined (equation (3)): y_{i,c,t+1} − y_{i,c,t} = α^y + β^y_o × SA_{i,c,t} + β^y_s × SAS_{i,c,t} + γ^y × X_{i,c,t} + ε_{i,c,t}
- Outcome variables: employment, revenues, investment, and labor productivity measured annually; monetary variables converted to 2015 constant U.S. dollars; capital variables deflated with World Bank WDI investment deflators.
- Robustness: controls for total debt (logs), working capital, sector-specific time trends; standard errors clustered at the firm level.

### Key empirical findings (selected)
- Own effects (Table 3):
  - Employment: State Aid 0.273* (0.159)
  - Revenue: State Aid 0.590* (0.311)
  - Investment: State Aid 0.440 (0.346) — not statistically significant
  - Productivity: State Aid -0.006 (0.726) — not statistically significant
  - Economic significance: one standard deviation shock = 1 percent excess return; estimated effects imply a 100 million euro unanticipated state aid (NPV) would lead the average firm to increase employment by 0.3 percent of 10,009 → 30 employees, and increase annual revenue by 0.6 percent of 2.45 billion euros → 15 million euros.
- Spillovers (Table 10 and joint Table 11):
  - State Aid Spillover: Employment -0.133* (0.071); Revenue -0.246** (0.117); Productivity -0.522** (0.256)
  - Joint specification: State Aid (own) still positive for employment and revenue, while State Aid Spillover negative and partially offsets own effects.
- Horizon and persistence (Table 12):
  - At 2-year horizon: own effects not significant; spillovers remain adverse and statistically significant.
  - State Aid (2 years later): Employment -0.081 (0.686); Revenue -0.126 (0.304)
  - State Aid Spillover (2 years later): Employment -0.208* (0.116); Revenue -0.460*** (0.177)
- Market concentration interaction (Table 13):
  - SAS X HHI: Employment -0.408* (0.231); Revenue -0.943** (0.391) — more adverse spillovers in more concentrated markets.

### Heterogeneity and mechanisms
- Firm size (Table 4):
  - Smaller firms benefit more:
    - Estimated employment increase for 10th percentile firms = 0.96 percent (statistically significant).
    - 90th percentile firms: -0.39 percent (not statistically significant).
  - Table 4 coefficients: State Aid: Employment 4.947** (2.198); Revenue 8.515** (4.106); Interaction State Aid X Size: Employment -0.233** (0.106); Revenue -0.393* (0.200).
- Debt-servicing capacity (Table 5):
  - More leveraged firms (lower ability to service debt) respond more.
  - State Aid X EBITDA-to-LongTermDebt: Employment -0.016*** (0.002); Revenue -0.036*** (0.002); Investment -0.031*** (0.006).
- Liquidity (Table 6):
  - Sensitivities decline with firm liquidity.
  - State Aid X Sales-to-Liabilities: Employment -0.141* (0.075); Revenue -0.575** (0.239); Investment -0.412* (0.230).
- Age (Table 8 and Table 9):
  - Younger firms increase employment and revenues more than older firms.
  - State Aid: Employment 0.654* (0.340); Revenue 2.063*** (0.689)
  - State Aid X Age: Employment -0.006* (0.003); Revenue -0.021*** (0.007)
  - Example: a one-year-old firm increases employment by 0.65 percent; a 40-year-old firm (average age) increases employment by 0.4 percent.
- R&D intensity (Table 7):
  - No significant differential response by R&D intensity.
  - Interaction State Aid X R&D Intensity: Employment -3.831 (6.173); Revenue -5.330 (6.430).
  - Interpretation: results point to financial-constraint channel rather than increased innovation.

### Robustness checks and sensitivity analyses
- Placebo test (Table 14):
  - Placebo using excess stock returns one week before announcements: coefficients not statistically different from zero.
  - Randomized State Aid: Employment 0.100 (0.110); Revenue 0.334 (0.255).
- Sample composition:
  - Dropping United Kingdom firms and including agricultural, financial, and real estate sectors: results similar (unreported).
- Alternative spillover definition:
  - Including same-country firms in spillover definition: results unchanged (not reported).
- Covid-19 interaction (Table 15):
  - Interaction of state aid with a Covid-19 dummy not significant for dependent variables; inclusion reduced overall significance due to short time dimension.
  - SA: Employment 0.191 (0.177); Revenue 0.418* (0.238); SA X I(Pandemic): Employment 0.133 (0.239).
- Aid type and additional controls:
  - Results similar for grants and loans; controls for total debt (logs), working capital, sector trends produced quantitatively similar results.
- Listed versus nonlisted firms (Table 16):
  - Spillovers of state aid from listed firms to other listed and nonlisted firms remain significantly negative for employment, revenues, and investment, but not for productivity.
  - State Aid Spillovers: Employment -0.264*** (0.029); Revenue -0.183*** (0.054); Investment -0.147*** (0.030).
- Robustness summary:
  - Multiple checks generally leave the main finding intact: negative spillovers to nonrecipient firms in the same sector are robust, especially for employment and revenue.

### Key contextual and sample statistics (exact figures)
- EU-wide state aid trend:
  - Increased from 0.5 percent of GDP in 2012 to nearly 1.5 percent of GDP in 2022; half of the latter is pandemic- and Russia-Ukraine war-related aid.
- Coverage of sample countries as of 2022:
  - The countries covered account for [85] percent of total reported state aid cases and 78 percent of the value of awarded state aid in the EU.
  - In 2022, state aid in the sample countries ranged from 1.1 percent of GDP in the Netherlands and Belgium to 1.8 percent of GDP in Germany.
- Sample firm size and representativeness:
  - Average revenues of firms in the sample: 2.4 billion euros.
  - Average employment: more than 10 thousand workers.
  - Largest firm in sample revenues: more than 341 billion euros.
  - In 2020 (last year including the United Kingdom), total firm value added in the sample varies between 5 percent of GDP in Italy and almost 20 percent of GDP in France and the United Kingdom.

### Data summary and definitions (selected exact figures and definitions)
- Table 1 (All firms, Observations = 7,408 unless otherwise noted)
  - State Aid: Mean 0.000; Median 0; Std. Deviation 0.010; Minimum -0.368; Maximum 0.155
  - State Aid Spillovers: Mean 0.000; Median 0; Std. Deviation 0.031; Minimum -0.337; Maximum 0.869
  - Employment: Mean 10009; Median 1035; Std. Deviation 33790; Minimum 20; Maximum 596452
  - Revenue: Mean 2446; Median 266.5; Std. Deviation 6286; Minimum 5.022; Maximum 48464
  - Investment: Observations 6,101; Mean 162.7; Median 3.796; Std. Deviation 2342.2; Minimum -18606; Maximum 137969
  - Productivity: Observations 6,883; Mean 141821; Median 92138; Std. Deviation 339622; Minimum 428.7; Maximum 18800000
  - Size (log of total assets): Mean 19.80; Median 19.6; Std. Deviation 2.160; Minimum 15.835; Maximum 26.08
  - Age: Mean 39.64; Median 27.5; Std. Deviation 36.90; Minimum 0.000; Maximum 273
  - R&D Intensity (obs = 1,827): Mean 0.056; Median 0.024; Std. Deviation 0.099; Minimum 0.000; Maximum 1.427
- Table 2 (Treated firms, Observations = 494 unless otherwise noted)
  - State Aid Shock: Mean 0.001; Median 0.000; Std. Deviation 0.053; Minimum -0.368; Maximum 0.450
  - State Aid Spillover: Mean 0.000; Median 0.000; Std. Deviation 0.003; Minimum -0.029; Maximum 0.028
  - Employment: Mean 24737; Median 2660; Std. Deviation 57005; Minimum 21; Maximum 385000
  - Revenue (A€C millions): Mean 8851; Median 807; Std. Deviation 21455; Minimum 1; Maximum 202692
  - Investment (A€C millions): Mean 288; Median 4; Std. Deviation 6949; Minimum -63022; Maximum 115321
  - Size (log of total assets): Mean 20.813; Median 20.760; Std. Deviation 2.553; Minimum 15.429; Maximum 26.776
  - Debt-to-Assets (obs = 493): Mean 0.923; Median 0.955; Std. Deviation 0.118; Minimum -0.465; Maximum 1.000
  - R&D Intensity (obs = 352): Mean 0.027; Median 0.001; Std. Deviation 0.075; Minimum 0.000; Maximum 0.803
  - Cash at Hand (A€C millions): Mean 0.293; Median 0.153; Std. Deviation 0.750; Minimum 0.000; Maximum 10.972
- Data definitions (exact):
  - State Aid Shock: "Percentage movement in a company’s stock price that exceeds the overall market movement on the official announcement date."
  - State Aid Spillover: "sum of state aid shocks in the same NACE2.0 4-digit sector but in different countries, adjusted by the size of the firms."
  - Employment: "Total number of employees included in the company’s payroll."
  - Investment: "Change in total assets between two periods."
  - Revenue: "Total operating revenues (Net sales + Other operating revenues+ Stock variations). The figures do not include VAT. Local differences may occur regarding excises taxes and similar obligatory payments for specific market of tobacco and alcoholic beverage industries.."
  - Productivity: "Value added per worker."
  - Size: "Total Assets = Fixed assets + Current assets."
  - Liquidity: "Sales Coverage Ratio = Sales / Liabilities."
  - Debt Servicing Capacity: "EBIDTA / Long-Term Debt."
  - Market Concentration (HHI): "Herfindahl–Hirschman index using revenues at the NACE 4-digit level."
- Note (exact): variables are based on Orbis with data cleaning and monetary variables converted to 2015 constant U.S. dollars; capital variables deflated with World Bank WDI investment deflators.

### Limitations and future research priorities
- Aggregating firm-level effects to macroeconomic and welfare outcomes is challenging and may miss general-equilibrium effects (e.g., wage increases from higher labor demand could dampen employment responses).
- Paper does not aggregate firm-level effects to macroeconomic/general-equilibrium welfare; quantifying overall welfare implications is beyond scope and flagged as a priority for future research.
- Priority for future work: comprehensive evaluation of state aid’s net impact on EU employment, investment, productivity, and overall welfare to inform industrial policy design that balances economic objectives with single market integrity.

*Source: Executive Summary of "A Bitter Aftertaste: How State Aid Affects Recipient Firms and Their Competitors in Europe" (December 11, 2024) — wpiea2024250-print-pdf - Executive Summary4*

### Executive Summary4

### wpiea2024250-print-pdf - Executive Summary4

### Executive summary — core findings
- State aid can temporarily boost firm-level employment and revenue of recipient firms but has no discernible impact on firm investment or productivity.
- Measured effects on recipient firms: employment increases by 0.3 percent and revenues increase by 0.6 percent in the following year.
- Adverse spillovers to competitor (nonrecipient) firms: decreases in employment, revenues, and productivity in the same sector; no detectable effect on investment for competitors.
- Magnitude and persistence of spillovers:
  - Adverse spillovers are at least half as large as the own-firm effects after one year.
  - Spillovers are more long-lasting and overturn the initial positive own-firm effects after two years.
- Heterogeneity of own-firm effects:
  - Stronger effects for firms that are smaller, have less liquidity, lower debt-servicing capacity, and are younger.
  - No differential response for firms with higher R&D intensity (no evidence that firms that spend more in R&D as a proportion of sales respond differently).
- Interpretation: state aid may overcome firm-level financial constraints but does not appear to enhance firms’ ability to innovate or raise productivity. Large-scale state aid programs would unlikely raise aggregate productivity in Europe given the negative spillovers.

### Policy implications and recommendations
- National-level state aid risks distorting competition within the EU single market and can create an uneven playing field that favors domestic firms over competitors from other member states.
- Net aggregate losses for the EU are possible because negative spillovers from state aid often surpass short-term benefits to recipients.
- Recommendation: when state aid is justified by well-identified externalities and less distortionary tools are absent, coordination at the EU level is preferable.
  - Pooling resources and competitively allocating aid across the Union could preserve market competition, encourage firm entry, and ensure a more efficient distribution of funds.
  - An EU-level approach can better capture gains across value chains spanning multiple member states, reduce capture by entrenched interests, and address disparities in fiscal capacity.
- An EU-coordinated industrial policy could help firms achieve economies of scale that would be difficult under fragmented national interventions, particularly amid increasing geoeconomic fragmentation.

### Research design and identification
- Scope and sample:
  - State aid awarded between 2016 and 2023 to listed nonfinancial firms in Belgium, France, Germany, Italy, the Netherlands, Spain, and the United Kingdom (until 2020).
  - State aid data from the European Commission’s State Aid Transparency Database; firm accounting and employment data from Orbis; excess equity returns (daily) from Bloomberg.
- Identification strategy:
  - High-frequency identification approach using excess equity returns of recipient firms on announcement dates to capture the nonanticipated component of state aid and address endogeneity (reverse causality and omitted variables) under market efficiency assumptions.
- Outcome variables:
  - Employment, revenues, investment, and labor productivity measured at the annual frequency; monetary variables converted to 2015 constant U.S. dollars; capital variables deflated with World Bank WDI investment deflators.
- Robustness and scope:
  - Results robust to various potential misspecifications.
  - Paper does not aggregate firm-level effects to macroeconomic/general-equilibrium welfare; quantifying overall welfare implications is beyond scope and flagged as a priority for future research.

### Key contextual and sample statistics (as reported)
- EU-wide state aid trend: increased from 0.5 percent of GDP in 2012 to nearly 1.5 percent of GDP in 2022; half of the latter is pandemic- and Russia-Ukraine war-related aid.
- Coverage of sample countries as of 2022:
  - The countries covered account for [85] percent of total reported state aid cases and 78 percent of the value of awarded state aid in the EU.
  - In 2022, state aid in the sample countries ranged from 1.1 percent of GDP in the Netherlands and Belgium to 1.8 percent of GDP in Germany.
- Sample firm size:
  - Average revenues of firms in the sample: 2.4 billion euros.
  - Average employment: more than 10 thousand workers.
  - Largest firm in sample revenues: more than 341 billion euros.
- Sample representativeness:
  - In 2020 (last year including the United Kingdom), total firm value added in the sample varies between 5 percent of GDP in Italy and almost 20 percent of GDP in France and the United Kingdom.

### Limitations and future research priorities
- Aggregating firm-level effects to macroeconomic and welfare outcomes is challenging and may miss general-equilibrium effects (e.g., wage increases from higher labor demand could dampen employment responses).
- Priority for future work: comprehensive evaluation of state aid’s net impact on EU employment, investment, productivity, and overall welfare to inform industrial policy design that balances economic objectives with single market integrity.

*Source: Executive Summary of "A Bitter Aftertaste: How State Aid Affects Recipient Firms and Their Competitors in Europe" (December 11, 2024) — wpiea2024250-print-pdf - Executive Summary4*

### 3.1  Identification of State Aid Shocks

### 3.1  Identification of State Aid Shocks

### Endogeneity challenge and identification logic
- State aid is endogenous: causality flows both ways between firm outcomes and aid; reverse causality arises because governments may assign aid to firms expected to perform poorly (picking losers) or well (choosing winners).
- Aggregate-data studies exacerbate endogeneity problems; microdata approaches are preferred for building more credible counterfactuals.
- The paper adapts an identification approach analogous to policy-shock identification in macroeconomics (Cochrane and Piazzesi, 2002; Bauer and Swanson, 2023) with three key assumptions:
  - Financial markets have the same information about state aid as the government and understand the government’s reaction function.
  - The systematic component of a policy measure is anticipated by financial markets and fully reflected in prices before the policy is announced.
  - Any asset price reaction observed after (but sufficiently close to) the announcement is a surprise and therefore exogenous; effects of systematic and non-systematic policy changes are assumed to be the same.
- Implication: a post-announcement equity return surprise identifies an exogenous variation in state aid.

### Measurement of the state aid shock
- Firm-specific state aid surprise is measured as:
  - the daily return on the stock of each firm in the sample (in excess of the average daily market return) on the date of granting the aid.
  - The date of granting is defined as the day when the legal right to receive the aid is conferred under the applicable national legal regime.
- The state aid shock:
  - is expressed in percent and can be positive (more aid than expected) or negative (less than anticipated).
  - is reported to be symmetric around zero, with most surprises being small.
- The state aid shock is reported as not correlated with observable firm characteristics (e.g., no discernible association with firm size), supporting exogeneity and a causal interpretation.

### State aid spillover variable
- For each firm, the state aid spillover (SAS) is defined as:
  - the average of state aid shocks (weighted by firm assets) in all other countries in a given year and in the same industry (NACE level 4).
- Intended interpretation: the effect on a given firm of state aid provided to other firms in the same industry.

### Advantages of this microdata surprise-identification approach
- Relies on microdata, which ameliorates reverse causality relative to aggregate studies.
- Refers to precise policy interventions: identifies time of intervention, type (mostly grants), and recipient; does not bundle different policy instruments over time.
- Short tracking window (seven years: 2016-2023) reduces problems from changing policy motivations (with the caveat that the Covid-19 crisis could be a watershed event; addressed in Section 4.3).

### Econometric specifications (benchmark)
- Three benchmark panel-difference specifications (with firm-level clustered standard errors):
  1. Own-effects (equation (1)):
     - y_{i,c,t+1} − y_{i,c,t} = α^y + β^y_o × SA_{i,c,t} + γ^y × X_{i,c,t} + ε_{i,c,t}
     - where y_{i,c,t+1} is log employment, log total revenue, investment in fixed assets (log change in total assets), or log productivity (value added per worker); SA is sum of state aid shocks for firm i in year t; X includes firm size (log total assets), year and firm fixed effects.
  2. Spillover-effects (equation (2)):
     - y_{i,c,t+1} − y_{i,c,t} = α^y + β^y_s × SAS_{i,c,t} + γ^y × X_{i,c,t} + ε_{i,c,t}
     - where SAS is the state aid spillover variable.
  3. Combined own and spillovers (equation (3)):
     - y_{i,c,t+1} − y_{i,c,t} = α^y + β^y_o × SA_{i,c,t} + β^y_s × SAS_{i,c,t} + γ^y × X_{i,c,t} + ε_{i,c,t}
- Robustness exercises: control for total debt (logs), working capital, sector-specific time trends.
- Standard errors clustered at the firm level.

### Key empirical findings (own effects)
- From equation (1) estimates (Table 3):
  - Firm employment and revenues improve after state aid is awarded.
  - Firm-level investment and productivity: effects not statistically significant.
  - Consistency with other studies: positive employment response but no productivity gain (Branstetter et al., 2023; Criscuolo et al., 2019).
- Magnitude of effects (economic significance described explicitly):
  - One standard deviation shock = 1 percent excess return.
  - Estimated effects: firm-level employment increases by 0.3 percent for treated firms; revenue increases by 0.6 percent.
  - Average firm market capitalization ≈ 7 billion euros; a 1.6 percent excess return ≈ a little over 100 million euro.
  - Interpretation: 100 million euros of unanticipated state aid (NPV) would lead the average firm to:
    - increase employment by 0.3 percent of 10,009 → 30 employees.
    - increase annual revenue by 0.6 percent of 2.45 billion euros → 15 million euros.
- Comparison with other studies:
  - Criscuolo et al. (2019) report a 4.7 percent employment increase for a one standard deviation increase in their measure vs. 0.3 percent here.
  - Possible explanations: their policies are more horizontal; average firm size here is much larger and small firms respond more.

### Heterogeneity: role of firm size and financial constraints
- Interaction specification (equation (4)):
  - y_{i,c,t+1} − y_{i,c,t} = α^y + β^y_1 × SA_{i,c,t} + β^y_2 × Size_{i,c,t} + β^y_3 × SA_{i,c,t} × Size_{i,c,t} + ε_{i,c,t}
- Results (Table 4):
  - Smaller firms benefit more from state aid:
    - Estimated employment increase for 10th percentile (smallest) firms = 0.96 percent (statistically significant).
    - 90th percentile (largest) firms: -0.39 percent (not statistically significant).
  - Revenue effect decreases with firm size; no size gradient for investment or productivity.
- Mechanisms discussed:
  - Smaller firms more likely financially constrained → larger response to aid.
  - Large firms may obtain aid through lobbying/rent-seeking without obligations.
  - Financial-constraint proxies are imperfect; size often negatively correlated with financial constraints.

### Additional heterogeneity: leverage, liquidity, age, and R&D intensity
- Interaction of state aid with debt-service capacity (EBITDA-to-long-term-debt) (Table 5):
  - More leveraged firms (lower ability to service debt) respond more to state aid.
  - Interaction is negative and statistically significant for employment, revenue, and investment; not significant for productivity.
- Interaction with liquidity (sales-to-total-liabilities) (Table 6):
  - Sensitivities of employment, revenues, and investment to state aid decrease with firm liquidity.
  - Productivity sensitivity remains negative and not statistically significant.
- Interaction with firm age (Table 8):
  - Younger firms increase employment and revenues more than older firms (but not investment or productivity).
  - Example: a one-year-old firm increases employment by 0.65 percent; a 40-year-old firm (average age) increases employment by 0.4 percent.
  - Firm age is also correlated with financial constraints; further checks undertaken.
- Interaction with R&D intensity (R&D spending to firm sales) (Table 7):
  - Firms with higher R&D intensity do not show differential employment, revenue, or sales responses to state aid.
  - Interpretation: significant interactions with age but not with R&D intensity suggest results capture financial-constraint mechanisms rather than greater propensity to innovate.

### Spillovers to competitors and market structure effects
- Spillover estimates (equation (2), Table 10):
  - State aid given to other firms in the same sector and different countries adversely affects non-recipient firms’ employment, revenue, and productivity.
  - Coefficients on SAS are statistically significant for all regressions except investment.
- Joint own- and spillover specification (equation (3)):
  - State aid increases employment and revenues for recipients, but adverse spillovers partially overturn these effects and are clearly negative for firm productivity.
- Horizon analysis:
  - At a 2-year horizon (Table 12): own effects are short-lived and not significant after two years; spillovers remain highly statistically significant and more adverse.
  - Cumulative own- and spillover effects (one- and two-year ahead): initially positive own effects are more than compensated by adverse spillovers.
- Market concentration interaction (Table 13):
  - Herfindahl-Hirschman index at NACE 4-digit level interacted with SAS shows:
    - Firms in less competitive (more concentrated) markets suffer greater adverse spillovers than firms in more competitive sectors.
  - Suggests industrial policy stands a better chance of positive impact in more competitive environments.

*Source: wpiea2024250-print-pdf - 3.1  Identification of State Aid Shocks*

### 4.3  Robustness

### 4.3  Robustness

### Placebo test
- Performed a placebo test replacing the state aid shock with the change in excess stock returns of the recipient firms one week before the state aid is announced.
- Rationale: unless announcements were anticipated (e.g., leaks), a placebo shock should have no effect on firm outcomes.
- Findings:
  - Results in Table 14 show that for all outcome variables, the coefficient of the placebo state aid shock is not statistically different than zero.
  - Interpretation: equity price movements on actual announcement days have predictive power for firm outcomes, while excess returns on other days do not—supporting the identification strategy.

### Sample composition checks
- Re-estimated equations (1)-(3) after dropping firms from the United Kingdom to check for Brexit or pre-Covid-19 (British firms ending in 2020) effects.
- Re-estimated including firms from the agricultural, financial, and real estate sectors to assess sample-selection concerns.
- Findings:
  - In both cases, the results (unreported) are similar to those reported in the main analysis.

### Alternative definition of spillovers
- Changed the definition of the state aid spillover measure to include firms in the same sector and country (instead of only firms from different countries) to check for unobservable country effects correlated with spillovers.
- Findings:
  - Results (not reported) were unchanged.

### Covid-19 interaction
- Added a double interaction of the state aid shock with a Covid-19 dummy in specification (1) to test whether pandemic-era extraordinary state aid drove results.
- Findings:
  - For all dependent variables, the interaction of state aid with the Covid-19 dummy was not significant (i.e., the effects of state aid were not different before and after Covid-19).
  - Due to the short time dimension of the data, the overall effect of state aid ceased to be significant when this interaction was included.

### Aid type and additional controls
- Estimated equation (1) separately for grants and loans (i.e., interest rate subsidies) using the same state aid shock definition.
- Controlled for total debt (in logs), working capital, and sector-specific time trends in separate robustness exercises.
- Findings:
  - Results were similar for both types of aid.
  - Quantitatively similar results obtained when adding the additional controls (available from the authors).

### Listed versus nonlisted firms
- Concern: state aid shock is defined only for listed firms (relies on excess equity returns), while listed firms differ from nonlisted firms (larger, older, more fixed assets, more leveraged, lower returns, different investment responsiveness).
- Approach:
  - Included both listed and nonlisted firms where possible.
  - Could not estimate specifications (1) and (3) for nonlisted firms because the state aid shock is undefined for them.
  - Estimated the spillover effects of state aid awarded to listed firms in the same sector for both listed and nonlisted firms (specification 2).
- Findings (Table 16):
  - Spillovers of state aid remain significantly negative for employment, revenues, and investment, but not for productivity.
  - Interpretation: evidence broadly supports robustness of findings, particularly for firm employment and revenue.

### Robustness summary
- Multiple robustness checks (placebo test, alternative samples, alternative spillover definitions, Covid-19 interactions, aid-type splits, additional controls, inclusion of nonlisted firms in spillover tests) generally leave the main findings qualitatively unchanged.
- Key consistent result: negative spillovers to nonrecipient firms in the same sector are robust, especially for employment and revenue.

*Source: wpiea2024250-print-pdf - 4.3  Robustness*

### References

### wpiea2024250-print-pdf - References

### Key references cited
- Abadie, A., S. Athey, G. W. Imbens, and J. M. Wooldridge (2022). When Should You Adjust Standard Errors for Clustering?*. The Quarterly Journal of Economics 138(1), 1–35.
- Aghion, P., J. Cai, M. Dewatripont, L. Du, A. Harrison, and P. Legros (2015). Industrial policy and competition. American Economic Journal: Macroeconomics 7(4).
- Bauer, M. D. and E. T. Swanson (2023). A reassessment of monetary policy surprises and high-frequency identification. NBER Macroeconomics Annual 37, 87–155.
- Branstetter, L. G., G. Li, and M. Ren (2023). Picking winners? government subsidies and firm productivity in china. Journal of Comparative Economics 51(4), 1186–1199.
- Criscuolo, C., R. Martin, H. G. Overman, and J. Van Reenen (2019). Some causal effects of an industrial policy. American Economic Review 109(1), 48–85.
- Dechezleprêtre, A., E. Einiö, R. Martin, K.-T. Nguyen, and J. Van Reenen (2023). Do tax incentives increase firm innovation? an rd design for r&d, patents, and spillovers. American Economic Journal: Economic Policy 15(4), 486–521.
- Juhász, R., N. Lane, and D. Rodrik (2024). The new economics of industrial policy. Annual Review of Economics.
- Tirole, J. (2010). The Theory of Corporate Finance. Princeton University Press.
- Tirole, J. (2018). Economics for the Common Good. Princeton University Press.
- Wei, S.-J., Z. Wei, and J. Xu (2021). On the market failure of “missing pioneers”. Journal of Development Economics 152, 102705.
- Additional working papers and IMF working papers cited (Evenett et al. 2024; Hodge et al. 2024; Mesquita Gabriel et al. 2022; Sraer and Thesmar 2018; others listed).

### Summary statistics (Tables 1–2) — selected exact figures
- Table 1 (All firms, Observations = 7,408 unless otherwise noted)
  - State Aid: Mean 0.000; Median 0; Std. Deviation 0.010; Minimum -0.368; Maximum 0.155
  - State Aid Spillovers: Mean 0.000; Median 0; Std. Deviation 0.031; Minimum -0.337; Maximum 0.869
  - Employment: Mean 10009; Median 1035; Std. Deviation 33790; Minimum 20; Maximum 596452
  - Revenue: Mean 2446; Median 266.5; Std. Deviation 6286; Minimum 5.022; Maximum 48464
  - Investment: Observations 6,101; Mean 162.7; Median 3.796; Std. Deviation 2342.2; Minimum -18606; Maximum 137969
  - Productivity: Observations 6,883; Mean 141821; Median 92138; Std. Deviation 339622; Minimum 428.7; Maximum 18800000
  - Size (log of total assets): Mean 19.80; Median 19.6; Std. Deviation 2.160; Minimum 15.835; Maximum 26.08
  - Age: Mean 39.64; Median 27.5; Std. Deviation 36.90; Minimum 0.000; Maximum 273
  - R&D Intensity (obs = 1,827): Mean 0.056; Median 0.024; Std. Deviation 0.099; Minimum 0.000; Maximum 1.427
- Table 2 (Treated firms, Observations = 494 unless otherwise noted)
  - State Aid Shock: Mean 0.001; Median 0.000; Std. Deviation 0.053; Minimum -0.368; Maximum 0.450
  - State Aid Spillover: Mean 0.000; Median 0.000; Std. Deviation 0.003; Minimum -0.029; Maximum 0.028
  - Employment: Mean 24737; Median 2660; Std. Deviation 57005; Minimum 21; Maximum 385000
  - Revenue (A€C millions): Mean 8851; Median 807; Std. Deviation 21455; Minimum 1; Maximum 202692
  - Investment (A€C millions): Mean 288; Median 4; Std. Deviation 6949; Minimum -63022; Maximum 115321
  - Size (log of total assets): Mean 20.813; Median 20.760; Std. Deviation 2.553; Minimum 15.429; Maximum 26.776
  - Debt-to-Assets (obs = 493): Mean 0.923; Median 0.955; Std. Deviation 0.118; Minimum -0.465; Maximum 1.000
  - R&D Intensity (obs = 352): Mean 0.027; Median 0.001; Std. Deviation 0.075; Minimum 0.000; Maximum 0.803
  - Cash at Hand (A€C millions): Mean 0.293; Median 0.153; Std. Deviation 0.750; Minimum 0.000; Maximum 10.972

### Main empirical findings (selected coefficients from Tables 3–16; standard errors in parentheses; significance markers preserved)
- Table 3: Effects of State Aid on Recipient Firms (N = 6004/6008/6008/5378; Within R2 = 0.044, 0.076, 0.009, 0.030)
  - Employment: State Aid 0.273* (0.159)
  - Revenue: State Aid 0.590* (0.311)
  - Investment: State Aid 0.440 (0.346)
  - Productivity: State Aid -0.006 (0.726)
  - Size coefficient: Employment 0.099*** (0.022); Revenue 0.053 (0.035); Investment 0.044 (0.027); Productivity -0.258*** (0.047)
- Table 4: Effects by Size
  - State Aid: Employment 4.947** (2.198); Revenue 8.515** (4.106); Investment 4.598 (5.607); Productivity 4.910 (10.479)
  - Interaction State Aid X Size: Employment -0.233** (0.106); Revenue -0.393* (0.200)
- Table 5: Effects by Debt-Servicing Capacity (N = 5,278/5,282/5,282/4,853)
  - State Aid: Employment 0.312* (0.163); Revenue 0.621* (0.318); Investment 0.560 (0.355)
  - State Aid X EBITDA-to-LongTermDebt: Employment -0.016*** (0.002); Revenue -0.036*** (0.002); Investment -0.031*** (0.006)
- Table 6: Effects by Liquidity (N = 5,010/5,025/5,025/4,476)
  - State Aid: Employment 0.638** (0.276); Revenue 2.243*** (0.818); Investment 1.619* (0.834)
  - State Aid X Sales-to-Liabilities: Employment -0.141* (0.075); Revenue -0.575** (0.239); Investment -0.412* (0.230)
- Table 7: Effects by R&D Intensity (N = 1,502/1,501/1,501/1,309)
  - State Aid: Employment 0.175 (0.230); Revenue 0.381* (0.226)
  - Interaction State Aid X R&D Intensity: Employment -3.831 (6.173); Revenue -5.330 (6.430)
- Table 8: Effects by Firm Age (N = 5,970/5,973/5,973/5,347)
  - State Aid: Employment 0.654* (0.340); Revenue 2.063*** (0.689); Investment 1.019 (0.771)
  - State Aid X Age: Employment -0.006* (0.003); Revenue -0.021*** (0.007)
- Table 9: Effects by Firm Age and Size (N = 5,970/5,973/5,973/5,347)
  - State Aid: Employment 4.666** (2.103); Revenue 7.288* (3.901)
  - State Aid X Size: Employment -0.206** (0.099)
  - State Aid X Age: Revenue -0.019*** (0.007)
- Table 10: Spillovers of State Aid on Other Firms (N = 6004/6008/6008/5378)
  - State Aid Spillover: Employment -0.133* (0.071); Revenue -0.246** (0.117); Productivity -0.522** (0.256)
- Table 11: Effects of State Aid on Firms and Spillovers (joint)
  - State Aid: Employment 0.268* (0.158); Revenue 0.580* (0.310)
  - State Aid Spillover: Employment -0.132* (0.072); Revenue -0.244** (0.117)
- Table 12: Effects with 2-Years Later (N = 4,637/4,647/4,674/4,092)
  - State Aid (2 years later): Employment -0.081 (0.686); Revenue -0.126 (0.304)
  - State Aid Spillover (2 years later): Employment -0.208* (0.116); Revenue -0.460*** (0.177)
- Table 13: Effects by Market Concentration (N = 6004/6008/6008/5378)
  - SA: Employment 0.270* (0.158); Revenue 0.578* (0.311)
  - SAS X HHI: Employment -0.408* (0.231); Revenue -0.943** (0.391)
- Table 14: Placebo Test (Randomized SA)
  - Randomized State Aid: Employment 0.100 (0.110); Revenue 0.334 (0.255)
  - Randomized State Aid Spillover coefficients all 0.000 (SEs: 0.001–0.002)
- Table 15: State Aid and Covid-19 Pandemic (N = 6004/6008/6008/5378)
  - SA: Employment 0.191 (0.177); Revenue 0.418* (0.238)
  - SA X I(Pandemic): Employment 0.133 (0.239); Revenue 0.291 (0.424); Investment 0.883* (0.476)
- Table 16: Spillovers to Other Listed and Nonlisted Firms (N = 289,372 / 430,924 / 430,924 / 198,427)
  - State Aid Spillovers: Employment -0.264*** (0.029); Revenue -0.183*** (0.054); Investment -0.147*** (0.030)

### Data definitions and sources (Table A1) — exact definitions preserved
- State Aid Shock: "Percentage movement in a company’s stock price that exceeds the overall market movement on the official announcement date."
- State Aid Spillover: "sum of state aid shocks in the same NACE2.0 4-digit sector but in different countries, adjusted by the size of the firms."
- Employment: "Total number of employees included in the company’s payroll."
- Investment: "Change in total assets between two periods."
- Sales: "Net sales."
- Revenue: "Total operating revenues (Net sales + Other operating revenues+ Stock variations). The figures do not include VAT. Local differences may occur regarding excises taxes and similar obligatory payments for specific market of tobacco and alcoholic beverage industries.."
- Value Added: "Profit for period + Depreciation + Taxation + Interests paid + Cost of employees."
- Productivity: "Value added per worker."
- Size: "Total Assets = Fixed assets + Current assets."
- Liquidity: "Sales Coverage Ratio = Sales / Liabilities."
- Debt Servicing Capacity: "EBIDTA / Long-Term Debt."
- Debt-to-Asset: "(Total assets - Issued Share Capital) / Total Assets."
- Cash: "Cash and cash equivalent."
- Age: "Number of years since firm’s inception."
- R & D Expenses: "Total amount of expenses on research and development activities."
- R&D Intensity: "Ratio of R & D expenses to sales."
- Market Concentration (HHI): "Herfindahl–Hirschman index using revenues at the NACE 4-digit level."

- Note (exact): "The variables are based on Orbis but differ from the original Orbis dataset due to various data cleaning processes. Additionally, the financial variables are deflated versions provided by this dataset. Non-capital monetary variables are converted from nominal to real variables, denominated in 2015 constant U.S. dollars, using deflators from various sources (OECD, Eurostat, CEIC database, government websites). For capital variables, we use the World Bank’s World Development Indicators (WDI) investment deflators at the country level."

*Source: wpiea2024250-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024250-print-pdf.pdf_