## Annex I.

## Source details

**Canonical URL:** [Annex I.](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024135-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024135-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024135-print-pdf.pdf.json)

---

### Introduction
- Carbon pricing increases the cost of carbon-based fuels relative to less carbon-intensive and/or energy efficient alternatives, creating incentives to switch technologies and behaviors.
- The impact of carbon pricing on firm performance depends on: the stringency of regulation, the level of the carbon price, availability of low-carbon alternatives, degree of competition, and ability to pass on costs to consumers (Dechezlepretre et al., 2023).
- Dataset constructed:
  - Roughly 1 million firms from 12 European countries over 1995-2020.
  - Subset of 1,870 firms that own at least one installation in the ETS between 2008 and 2020.
  - Combines firm-level financial accounts from Bureau van Dijk’s Orbis with: OECD’s Environmental Policy Stringency (EPS) Index (market-based component), EU ETS emissions allowance trading prices, and information on whether firms have payable emissions in a given year.

### Major empirical approach
- Panel regression specification exploiting:
  - Heterogeneity by firm type (regulated under the ETS or not).
  - Heterogeneity in carbon price exposure measured by trading price of emissions permits in the EU ETS market.
  - Market-based component of OECD EPS index as the measure of tightening price-based regulations.

### Key findings on firm performance
- Average firm:
  - More stringent market-based climate policies have no sizable impact on economic performance during periods of low carbon prices.
- Non-ETS firms during periods of high carbon prices:
  - Face higher input costs.
  - Also earn more revenues, leaving profitability stable (implying changes in output composition or cost pass-through to customers).
- ETS firms, heterogeneity by payable emissions status:
  - Firms with no payable emissions:
    - Experience windfall profits during periods of high carbon prices when climate policies become more stringent.
    - Likely mechanism: ability to sell unused emissions allowances in secondary markets at higher prices.
  - Firms with positive payable emissions:
    - Face heightened input costs amid tighter regulations.
    - Also exhibit increased turnover.
    - There is no statistically significant impact on overall profitability in some specifications; other specifications show negative impacts (see coefficient rows).
    - These firms appear to protect profitability by adjusting production mix or passing costs to consumers.
    - An increase in the market-based EPS index by one unit is associated with an increase in fixed assets by about 4.5 percentage points among these firms (text statement).
  - Investment composition:
    - Relative to firms without payable emissions, firms with payable emissions redirect investment towards greater investment in intangible capital.
    - Prior literature links investments in production efficiency and energy efficiency to changes in intangible capital (Wu and Wang, 2022; Yang and Shi, 2018).
- Timing and drivers:
  - Regulated firms with payable emissions respond differently in periods of high ETS price levels and not necessarily high price volatility, suggesting impacts are driven more by binding budget constraints than price fluctuations.

### Numeric and temporal highlights
- Dataset span: 1995-2020.
- ETS-regulated firm subset: 1,870 firms (2008–2020 coverage for ETS installations).
- Countries covered in average EPS calculations: 12 European countries.
- EPS index trend:
  - Stringency remained low until 2012 with large year-to-year variation.
  - Starting from 2012, average market-based stringency has almost doubled by 2020.
- EU ETS phases and details:
  - Phase 1: 2005-2007 (trial period, free allowances).
  - Phase 2: 2008-2012 (cap reduced by 6.5% and added Iceland, Liechtenstein and Norway).
  - Phase 3: 2013-2020 (cap tightened by 1.74% annually).
  - Phase 4: 2021-2030 (cap tightened by 2.2% annually, scope expanded to include aviation).
- ETS price history:
  - Around €8 in 2005.
  - Around €30 in 2006.
  - Fell to almost zero in 2007.
  - Around €14 in Phase 2.
  - Around €11 in Phase 3.
  - Surged to above €70 in the first 2.5 years of Phase 4.
- Policy target: net GHG emissions reduced by at least 55% by 2030 compared to 1990 levels (European Green Deal objective).
- Reported total effect calculation note:
  - Total effect is calculated as a linear combination of the marginal effects for stricter climate regulations, high carbon price periods and ETS firms with payable emissions.

### Data patterns and distributional features
- ETS sectors vs non-ETS sectors (Nace Rev. 2, 4-digit classification) — descriptive patterns:
  - Despite higher median investment levels, ETS sectors are less productive.
  - Within ETS sectors, ETS-reporting firms are even less productive despite even higher investment levels.
  - ETS sectors are larger in median employment and asset size; ETS-reporting firms are larger still.
  - Size variables, particularly employment, have a wider distributional range among ETS firms.
- Distributional irregularities:
  - Coefficient of variation of firms above the 3rd quartile of TFP distribution is nearly two times higher among ETS firms than among firms in non-ETS sectors.
  - For fixed investment, this metric is 15% higher for ETS firms.

### Robustness and methodological checks
- Robustness exercises:
  - Quantile panel model estimated for the median of the distribution of outcome variables to account for outliers and distributional irregularities; main effects persist.
  - Results robust to introduction of additional firm-level control variables.
  - Results robust when using an alternative measure of tangible fixed assets.
- Interpretation:
  - Findings align with literature that finds no evidence that more stringent climate policies harmed firms’ economic performance (Dechezleprêtre et al. 2023; Dechezleprêtre and Krause 2022).
  - Adds evidence that tighter climate policies are associated with increases in fixed assets of ETS-regulated firms during elevated carbon price periods over 2012-2020.

### Contribution to literature
- Extends prior work by:
  - Using a large cross-country firm-level database to assess direct impacts on regulated firms and spillovers on non-regulated firms in the same country and sector.
  - Leveraging variations in emissions permit prices to test differential impacts during high and low price periods.
  - Providing robustness across the distribution of ETS firms and multiple outcome variables.
- Links to prior empirical findings:
  - Confirms and extends findings of increased investment in fixed capital among regulated firms (Marin et al. 2018; Dechezleprêtre et al. 2023).
  - Complements studies on heterogeneous firm responses to climate policy and ETS price changes (Bijnens and Swartenbroekx, 2022; Berthold et al., 2023).

### Data and empirical specification (section 3) — dataset and key variables
- Novel firm-level dataset matching financial accounts across EU economies with country- and sector-wide regulatory and energy variables.
- Environmental Policy Stringency (EPS) index from OECD used to identify firm exposure at the country-year level. EPS composed of sub-indices capturing market-based (MB) and non-market-based (NMB) policies.
  - Market-based policies include emissions trading schemes, CO2 taxes, taxes on NOx, SOx and fuel taxes.
  - Non-market-based policies include emission limits and content limits.
- Market-based EPS index: twofold increase for the median country between 2005 and 2020; starting from 2012, market-based EPS indices rose in all countries by 0.9 points on average.
  - France tightened by 3.2 points since 2012; Belgium and Portugal by 1.2 points since 2012.
  - Greece and Germany tightened by 0.5 points since 2012 but remained at the bottom of distribution.
- Firm financials from Bureau van Dijk’s Orbis: turnover, cost of goods sold (material costs), stock of fixed assets, EBITDA.
  - Flow variables converted to real variables using 2-digit NACE level deflators and World Bank WDI investment deflators at country level (for fixed assets).
  - Sample restricted to non-financial firms with at least 10 consecutive years of sales data.
- EU ETS data on verified emissions and free allocated emission allowances matched to Orbis firms with ETS-regulated installations.
  - Orbis identifiers available for about 7,200 ETS-regulated firms, covering 66 percent of all installations.

### Sample coverage and summary statistics (section 3)
- Baseline panel:
  - 980 thousand firms (including 1,870 ETS-regulated firms) from 12 European countries over 1995-2020.
  - Countries: Belgium, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Italy, Portugal, Spain, United Kingdom. Sample period coverage varies by country.
- Selected summary statistics (Full sample / EU ETS sample):
  - Full sample (Obs = 8,815,354 unless noted):
    - Turnover (net change): Obs 8,815,354; Mean 0.020; Std. dev. 0.280; Min -11.956; Max 13.816
    - Material costs (net change): Obs 7,387,437; Mean 0.008; Std. dev. 0.562; Min -14.411; Max 15.126
    - EBITDA (net change): Obs 8,815,354; Mean 0.005; Std. dev. 0.825; Min -13.448; Max 11.365
    - Fixed assets (net change): Obs 8,815,354; Mean 0.028; Std. dev. 0.544; Min -15.585; Max 13.455
    - N. of employees (log): Obs 8,815,354; Mean 2.21; Std. dev. 1.44; Min 0.00; Max 13.42
    - Total assets (log): Obs 8,815,354; Mean 13.88; Std. dev. 1.83; Min 2.75; Max 27.08
  - EU ETS subsample (Obs = 6,356 unless noted):
    - Turnover (net change): Obs 6,356; Mean 0.012; Std. dev. 0.209; Min -6.817; Max 4.034
    - Material costs (net change): Obs 5,540; Mean 0.005; Std. dev. 0.373; Min -6.925; Max 7.008
    - EBITDA (net change): Obs 6,356; Mean 0.026; Std. dev. 0.576; Min -6.743; Max 6.927
    - Fixed assets (net change): Obs 6,356; Mean 0.020; Std. dev. 0.291; Min -9.694; Max 9.729
    - N. of employees (log): Obs 6,356; Mean 5.65; Std. dev. 1.76; Min 0.00; Max 13.42
    - Total assets (log): Obs 6,356; Mean 18.89; Std. dev. 1.82; Min 13.73; Max 26.98

### Empirical strategy (baseline specification) — key model elements
- Baseline panel regression (equation (1)):
  - Dependent variables: ∆y_i c t = y_i c t − y_i c t−1, where y is log of turnover, input costs (material costs), profits (EBITDA), fixed assets (total non-current assets after depreciation).
  - Key regressors: changes in market-based and non-market-based EPS (∆EPS Market, ∆EPS Non-Market) and interactions with firm ETS status and high carbon price indicators.
  - X_i c t vector includes:
    - 1,
    - D_i CCA2 (sector dummy),
    - ETS_i0 (dummy for firm with at least one ETS installation),
    - ETS0 × D_i CCA2,
    - ETS_i P (indicator for positive payable emissions: emissions exceeding free allowance),
    - ETS_i P × D_i CCA2
  - D_i CCA2: binary indicating if carbon prices are above the 75th percentile calculated over 2008-2020 (excluding 2005-2007).
  - Controls: firm-level controls W and country-level controls Z (including ∆EPS Non-Market), country-sector (D_ic) and sector-year (D_ci) fixed effects, firm fixed effects D_i; firm-level fixed effects absorbed.
- Robust standard errors clustered at the firm level.

### Robustness: distributional and quantile approach
- Quantile regression at the median (q = 0.5) using Mundlak (1978) device to control for unobserved firm-level heterogeneity in pooled quantile regressions (equation (2)).
  - V_i c t−1 = [∆EPS Market × X_i c t, W_i c t, Z_i c t]; V̄_i c are firm-specific time means.
  - In empirical application, due to convergence difficulties, skip W and Z in V̄; V̄ includes means of ∆EPS Market × X variables.
  - Firm-clustered standard errors used.

### Selected empirical estimates and interpretation (summary of Table 2 and figures)
- Lag ∆EPS Market:
  - ∆ Log Input Costs −0.00989***; ∆ Log Turnover −0.0131***; ∆ Log Profits 0.0104***; ∆ Log Fixed Assets 0.00296**
- Lag ∆EPS Market × High carbon price:
  - ∆ Log Input Costs 0.0503***; ∆ Log Turnover 0.0537***; ∆ Log Profits 0.0159***; ∆ Log Fixed Assets -0.000477
- Lag ∆EPS Market × ETS firm:
  - ∆ Log Input Costs −0.0346***; ∆ Log Turnover −0.0114*; ∆ Log Profits −0.0448**; ∆ Log Fixed Assets −0.0113
- Lag ∆EPS Market × ETS firm × High carbon price:
  - ∆ Log Input Costs −0.0474*; ∆ Log Turnover −0.0292**; ∆ Log Profits 0.150***; ∆ Log Fixed Assets −0.0176
- Lag ∆EPS Market × ETS firm with payable emissions × High carbon price:
  - ∆ Log Input Costs 0.0973**; ∆ Log Turnover 0.0435*; ∆ Log Profits −0.217***; ∆ Log Fixed Assets 0.0598*

- Interpretation of heterogeneous effects (from figures and text):
  - ETS-regulated firms overall: decrease in turnover by 2.4 percentage points, profits unchanged on average, no significant change in fixed assets for average ETS firm (text statement).
  - ETS firms without payable emissions during high carbon price periods:
    - Declines in input costs and large increases in profitability (driven by ability to trade surplus allowances).
  - ETS firms with payable emissions during high carbon price periods:
    - Increase in input costs of over 6 percent (text statement).
    - Turnover increases by 3 percent (text statement).
    - Profitability impacts negative but Table 2 triple interaction shows ∆ Log Profits −0.217***.
    - Fixed assets: tightening by one unit associated with a rise in fixed assets by about 4.5 percentage points when carbon prices are high (text statement).
  - Investment composition:
    - ETS firms with payable emissions in periods of high carbon prices: increases in total fixed assets and tangible fixed assets, implying intangible fixed asset investments also increasing.
    - ETS firms without payable emissions in periods of high carbon prices: only tangible fixed assets increase (7 percentage points); total fixed assets remain unchanged, implying a decline in investment in intangibles.

### Robustness (section 5) — checks and main robustness findings
- Robustness checks performed:
  - Controlled for additional firm-level characteristics: cash-to-asset ratio and leverage (debt-to-asset ratio).
  - Verified sensitivity to alternative measure of investment: tangible fixed assets.
  - Explored heterogeneity by carbon price environment: high vs low carbon prices and high carbon price volatility.
  - Estimated quantile regressions (median, q = 0.5).
  - Sample and data: panel regressions for 12 European countries covering 1995–2020.
- Main robustness findings:
  - Inclusion of (lagged) cash-to-asset and debt-to-asset ratios: main results remain robust (Annex Table 1).
  - Estimated impacts remain broadly the same across ETS-regulated firm subgroups, including during periods of high carbon prices.
  - Precision of some parameter estimates improves with additional controls.
  - Difference from baseline: more stringent climate policies are associated with slightly larger increase in fixed assets (by 6 percentage points compared to 4.5 in the baseline).
- High vs low carbon price environments:
  - When carbon prices are low:
    - No change in input costs or profits of ETS firms.
    - ETS firms with payable emissions: tightening of EPS when carbon prices are low is associated with almost 9 percent decline in input costs and a 7 percent decline in turnover.
    - More stringent climate policies do not enhance investment by ETS firms with payable emissions in a low carbon price environment.
  - When carbon prices are high:
    - ETS firms with payable emissions increase investment levels during periods of high carbon prices (confirmed in both average and median analyses).
- Carbon price volatility:
  - During periods of high carbon price volatility:
    - No significant response of ETS firms with payable emissions to tightening of climate policies.
    - Higher volatility does not induce ETS firms to accumulate fixed assets.
    - ETS firms with no payable emissions reduce their fixed assets on average despite benefiting from volatile carbon prices.
- Quantile (median) regression results:
  - Median effects (q = 0.5) largely not statistically different from baseline results at the 5% significance level.
  - Median effects for input costs and turnover become modestly negative among ETS firms with payable emissions in periods of high carbon prices.
  - Median profits of ETS firms remain protected or even increase in periods of high carbon prices.
  - Investment impact among ETS firms with payable emissions: median effects roughly 3%, modestly weaker than average results but still statistically significant.

### Numerical robustness table highlights (Annex Table 1 and Annex Table 2 extracts)
- Lag ∆EPS Market row (selected): -0.00989*** -0.0106*** -0.0131*** -0.0145*** 0.0104*** 0.0108*** 0.00296** -0.000390 0.00713*** 0.00525***
- Lag ∆EPS Market × High carbon price row (selected): 0.0503*** 0.0530*** 0.0537*** 0.0576*** 0.0160*** 0.0126** -0.000476 0.0180*** 0.00867** 0.0221***
- Lag ∆EPS Market × ETS firm row (selected): -0.0338*** -0.0308** -0.0113* -0.0121* -0.0461** -0.0486** -0.0113 -0.0103 -0.0157* -0.0143
- Lag ∆EPS Market × ETS firm × High carbon price row (selected): -0.0487** -0.0573** -0.0293** -0.0385*** 0.154*** 0.133*** -0.0163 -0.00183 0.0748*** 0.0976***
- Lag ∆EPS Market × ETS firm with payable emissions × High carbon price row (selected): 0.101** 0.114*** 0.0435* 0.0524** -0.228*** -0.187** 0.0559* 0.0340 -0.0292 -0.0485
- Lag ∆EPS Non-Market row (selected): 0.0109*** 0.0100*** 0.00169*** 0.00169*** -0.00408*** -0.00701*** -0.00490*** -0.00242*** -0.00587*** -0.00293***
- Lag Log Employment row (selected): -0.0538*** -0.0480*** -0.0523*** -0.0468*** 0.000802 0.00384*** 0.0184*** 0.0191*** 0.00825*** 0.00816***
- Lag Log Total Assets row (selected): -0.130*** -0.137*** -0.112*** -0.116*** -0.245*** -0.260*** -0.187*** -0.169*** -0.171*** -0.155***
- Observations row examples: 7,326,460 6,223,366 7,326,460 6,223,366 7,326,460 6,223,366 7,326,460 6,223,366 7,326,460 6,236,517 6,165 198
- R-squared examples: 0.126 0.138 0.201 0.214 0.091 0.109 0.135 0.165 0.134 0.157

### Implications highlighted in robustness section
- More stringent market-based climate policies do not negatively impact the average firm or ETS firms; they incentivize investment in low-carbon technologies and emission reductions.
- Cost pass-through is limited for most firms; only a few firms pass increased input costs downstream.
- Policy design and coordination matter to avoid fragmentation and to address distributional effects across firms (especially hard-to-abate sectors).
- Future research directions suggested: structural analysis of regulatory shock transmission, role of firm innovation capacity and distance to technological frontier in investment responses, and quantile regression analysis beyond the median to assess tail risks.

*International Monetary Fund — Annex I to "Firms’ Response to Climate Regulations" (working paper excerpt).*

### Annex I. ...............................................................................................................

### Annex I.

### Introduction
- Carbon pricing increases the cost of carbon-based fuels relative to less carbon-intensive and/or energy efficient alternatives, creating incentives to switch technologies and behaviors.
- The impact of carbon pricing on firm performance depends on: the stringency of regulation, the level of the carbon price, availability of low-carbon alternatives, degree of competition, and ability to pass on costs to consumers (Dechezlepretre et al., 2023).
- Dataset constructed:
  - Roughly 1 million firms from 12 European countries over 1995-2020.
  - Subset of 1,870 firms that own at least one installation in the ETS between 2008 and 2020.
  - Combines firm-level financial accounts from Bureau van Dijk’s Orbis with: OECD’s Environmental Policy Stringency (EPS) Index (market-based component), EU ETS emissions allowance trading prices, and information on whether firms have payable emissions in a given year.

### Major empirical approach
- Panel regression specification exploiting:
  - Heterogeneity by firm type (regulated under the ETS or not).
  - Heterogeneity in carbon price exposure measured by trading price of emissions permits in the EU ETS market.
  - Market-based component of OECD EPS index as the measure of tightening price-based regulations.

### Key findings on firm performance
- Average firm:
  - More stringent market-based climate policies have no sizable impact on economic performance during periods of low carbon prices.
- Non-ETS firms during periods of high carbon prices:
  - Face higher input costs.
  - Also earn more revenues, leaving profitability stable (implying changes in output composition or cost pass-through to customers).
- ETS firms, heterogeneity by payable emissions status:
  - Firms with no payable emissions:
    - Experience windfall profits during periods of high carbon prices when climate policies become more stringent.
    - Likely mechanism: ability to sell unused emissions allowances in secondary markets at higher prices.
  - Firms with positive payable emissions:
    - Face heightened input costs amid tighter regulations.
    - Also exhibit increased turnover.
    - There is no statistically significant impact on overall profitability.
    - These firms appear to protect profitability by adjusting production mix or passing costs to consumers.
    - An increase in the market-based EPS index by one unit is associated with an increase in fixed assets by about 4.5 percentage points among these firms.
  - Investment composition:
    - Relative to firms without payable emissions, firms with payable emissions redirect investment towards greater investment in intangible capital.
    - Prior literature links investments in production efficiency and energy efficiency to changes in intangible capital (Wu and Wang, 2022; Yang and Shi, 2018).
- Timing and drivers:
  - Regulated firms with payable emissions respond differently in periods of high ETS price levels and not necessarily high price volatility, suggesting impacts are driven more by binding budget constraints than price fluctuations.

### Numeric and temporal highlights
- Dataset span: 1995-2020.
- ETS-regulated firm subset: 1,870 firms (2008–2020 coverage for ETS installations).
- Countries covered in average EPS calculations: 12 European countries.
- EPS index trend:
  - Stringency remained low until 2012 with large year-to-year variation.
  - Starting from 2012, average market-based stringency has almost doubled by 2020.
- EU ETS phases and details:
  - Phase 1: 2005-2007 (trial period, free allowances).
  - Phase 2: 2008-2012 (cap reduced by 6.5% and added Iceland, Liechtenstein and Norway).
  - Phase 3: 2013-2020 (cap tightened by 1.74% annually).
  - Phase 4: 2021-2030 (cap tightened by 2.2% annually, scope expanded to include aviation).
- ETS price history:
  - Around €8 in 2005.
  - Around €30 in 2006.
  - Fell to almost zero in 2007.
  - Around €14 in Phase 2.
  - Around €11 in Phase 3.
  - Surged to above €70 in the first 2.5 years of Phase 4.
- Policy target: net GHG emissions reduced by at least 55% by 2030 compared to 1990 levels (European Green Deal objective).
- Reported total effect calculation note:
  - Total effect is calculated as a linear combination of the marginal effects for stricter climate regulations, high carbon price periods and ETS firms with payable emissions.

### Data patterns and distributional features
- ETS sectors vs non-ETS sectors (Nace Rev. 2, 4-digit classification) — descriptive patterns:
  - Despite higher median investment levels, ETS sectors are less productive.
  - Within ETS sectors, ETS-reporting firms are even less productive despite even higher investment levels.
  - ETS sectors are larger in median employment and asset size; ETS-reporting firms are larger still.
  - Size variables, particularly employment, have a wider distributional range among ETS firms.
- Distributional irregularities:
  - Coefficient of variation of firms above the 3rd quartile of TFP distribution is nearly two times higher among ETS firms than among firms in non-ETS sectors.
  - For fixed investment, this metric is 15% higher for ETS firms.

### Robustness and methodological checks
- Robustness exercises:
  - Quantile panel model estimated for the median of the distribution of outcome variables to account for outliers and distributional irregularities; main effects persist.
  - Results robust to introduction of additional firm-level control variables.
  - Results robust when using an alternative measure of tangible fixed assets.
- Interpretation:
  - Findings align with literature that finds no evidence that more stringent climate policies harmed firms’ economic performance (Dechezleprêtre et al. 2023; Dechezleprêtre and Krause 2022).
  - Adds evidence that tighter climate policies are associated with increases in fixed assets of ETS-regulated firms during elevated carbon price periods over 2012-2020.

### Contribution to literature
- Extends prior work by:
  - Using a large cross-country firm-level database to assess direct impacts on regulated firms and spillovers on non-regulated firms in the same country and sector.
  - Leveraging variations in emissions permit prices to test differential impacts during high and low price periods.
  - Providing robustness across the distribution of ETS firms and multiple outcome variables.
- Links to prior empirical findings:
  - Confirms and extends findings of increased investment in fixed capital among regulated firms (Marin et al. 2018; Dechezleprêtre et al. 2023).
  - Complements studies on heterogeneous firm responses to climate policy and ETS price changes (Bijnens and Swartenbroekx, 2022; Berthold et al., 2023).

_International Monetary Fund — Annex I to "Firms’ Response to Climate Regulations" (working paper excerpt)._

### 3.  Data and Empirical Specification

### 3.  Data and Empirical Specification

### Dataset and key variables
- Novel firm-level dataset matching financial accounts across EU economies with country- and sector-wide regulatory and energy variables.
- Environmental Policy Stringency (EPS) index from OECD used to identify firm exposure at the country-year level. EPS composed of sub-indices capturing market-based (MB) and non-market-based (NMB) policies.
  - Market-based policies include emissions trading schemes, CO2 taxes, taxes on NOx, SOx and fuel taxes.
  - Non-market-based policies include emission limits and content limits.
- Market-based EPS index: twofold increase for the median country between 2005 and 2020; starting from 2012, market-based EPS indices rose in all countries by 0.9 points on average.
  - France tightened by 3.2 points since 2012; Belgium and Portugal by 1.2 points since 2012.
  - Greece and Germany tightened by 0.5 points since 2012 but remained at the bottom of distribution.
- Firm financials from Bureau van Dijk’s Orbis: turnover, cost of goods sold (material costs), stock of fixed assets, EBITDA.
  - Flow variables converted to real variables using 2-digit NACE level deflators and World Bank WDI investment deflators at country level (for fixed assets).
  - Sample restricted to non-financial firms with at least 10 consecutive years of sales data.
- EU ETS data on verified emissions and free allocated emission allowances matched to Orbis firms with ETS-regulated installations.
  - Orbis identifiers available for about 7,200 ETS-regulated firms, covering 66 percent of all installations.

### Sample coverage and summary statistics
- Baseline panel:
  - 980 thousand firms (including 1,870 ETS-regulated firms) from 12 European countries over 1995-2020.
  - Countries: Belgium, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Italy, Portugal, Spain, United Kingdom. Sample period coverage varies by country.
- Table 1 summary statistics (Full sample / EU ETS sample):
  - Full sample (Obs = 8,815,354 unless noted):
    - Turnover (net change): Obs 8,815,354; Mean 0.020; Std. dev. 0.280; Min -11.956; Max 13.816
    - Material costs (net change): Obs 7,387,437; Mean 0.008; Std. dev. 0.562; Min -14.411; Max 15.126
    - EBITDA (net change): Obs 8,815,354; Mean 0.005; Std. dev. 0.825; Min -13.448; Max 11.365
    - Fixed assets (net change): Obs 8,815,354; Mean 0.028; Std. dev. 0.544; Min -15.585; Max 13.455
    - N. of employees (log): Obs 8,815,354; Mean 2.21; Std. dev. 1.44; Min 0.00; Max 13.42
    - Total assets (log): Obs 8,815,354; Mean 13.88; Std. dev. 1.83; Min 2.75; Max 27.08
  - EU ETS subsample (Obs = 6,356 unless noted):
    - Turnover (net change): Obs 6,356; Mean 0.012; Std. dev. 0.209; Min -6.817; Max 4.034
    - Material costs (net change): Obs 5,540; Mean 0.005; Std. dev. 0.373; Min -6.925; Max 7.008
    - EBITDA (net change): Obs 6,356; Mean 0.026; Std. dev. 0.576; Min -6.743; Max 6.927
    - Fixed assets (net change): Obs 6,356; Mean 0.020; Std. dev. 0.291; Min -9.694; Max 9.729
    - N. of employees (log): Obs 6,356; Mean 5.65; Std. dev. 1.76; Min 0.00; Max 13.42
    - Total assets (log): Obs 6,356; Mean 18.89; Std. dev. 1.82; Min 13.73; Max 26.98

### Empirical strategy (baseline specification)
- Baseline panel regression (equation (1)):
  - Dependent variables: ∆y_i c t = y_i c t − y_i c t−1, where y is log of turnover, input costs (material costs), profits (EBITDA), fixed assets (total non-current assets after depreciation).
  - Key regressors: changes in market-based and non-market-based EPS (∆EPS Market, ∆EPS Non-Market) and interactions with firm ETS status and high carbon price indicators.
  - X_i c t vector includes:
    - 1,
    - D_i CCA2 (sector dummy),
    - ETS_i0 (dummy for firm with at least one ETS installation),
    - ETS0 × D_i CCA2,
    - ETS_i P (indicator for positive payable emissions: emissions exceeding free allowance),
    - ETS_i P × D_i CCA2
  - D_i CCA2: binary indicating if carbon prices are above the 75th percentile calculated over 2008-2020 (excluding 2005-2007).
  - Controls: firm-level controls W and country-level controls Z (including ∆EPS Non-Market), country-sector (D_ic) and sector-year (D_ci) fixed effects, firm fixed effects D_i; firm-level fixed effects absorbed.
- Robust standard errors clustered at the firm level.

### Robustness: distributional and quantile approach
- Quantile regression at the median (q = 0.5) using Mundlak (1978) device to control for unobserved firm-level heterogeneity in pooled quantile regressions (equation (2)).
  - V_i c t−1 = [∆EPS Market × X_i c t, W_i c t, Z_i c t]; V̄_i c are firm-specific time means.
  - In empirical application, due to convergence difficulties, skip W and Z in V̄; V̄ includes means of ∆EPS Market × X variables.
  - Firm-clustered standard errors used.

### Key empirical findings (summary of Table 2 and figures)
- Aggregate / non-ETS firms:
  - More stringent market-based climate policies have had no significant impact on performance of an average non-ETS firm.
  - An increase in the market-based EPS index by one unit is associated with:
    - Lag ∆EPS Market: ∆ Log Input Costs −0.00989***; ∆ Log Turnover −0.0131***; ∆ Log Profits 0.0104***; ∆ Log Fixed Assets 0.00296**
  - Interpretation note: an average increase in the market-based EPS index for 12 countries over 2012–20 is about 0.9; a unit increase can be interpreted as tightening over the medium term.
- Role of carbon price and ETS status (selected interaction estimates from Table 2):
  - Lag ∆EPS Market × High carbon price:
    - ∆ Log Input Costs 0.0503***; ∆ Log Turnover 0.0537***; ∆ Log Profits 0.0159***; ∆ Log Fixed Assets -0.000477
  - Lag ∆EPS Market × ETS firm:
    - ∆ Log Input Costs −0.0346***; ∆ Log Turnover −0.0114*; ∆ Log Profits −0.0448**; ∆ Log Fixed Assets −0.0113
  - Lag ∆EPS Market × ETS firm × High carbon price:
    - ∆ Log Input Costs −0.0474*; ∆ Log Turnover −0.0292**; ∆ Log Profits 0.150***; ∆ Log Fixed Assets −0.0176
  - Lag ∆EPS Market × ETS firm with payable emissions × High carbon price:
    - ∆ Log Input Costs 0.0973**; ∆ Log Turnover 0.0435*; ∆ Log Profits −0.217***; ∆ Log Fixed Assets 0.0598*
- Interpretation of heterogeneous effects (from Figures 4 and 5 and accompanying text):
  - ETS-regulated firms overall: decrease in turnover by 2.4 percentage points, profits unchanged on average, no significant change in fixed assets for average ETS firm.
  - ETS firms without payable emissions during high carbon price periods:
    - Report declines in input costs and large increases in profitability (driven by ability to trade surplus allowances).
  - ETS firms with payable emissions during high carbon price periods:
    - Increase in input costs of over 6 percent (text statement).
    - Turnover increases by 3 percent (text statement), indicating some pass-through of higher carbon prices.
    - Profitability impacts negative but not statistically significant (text statement); Table 2 shows ∆ Log Profits −0.217*** in the triple interaction (significant negative).
    - Fixed assets: tightening by one unit associated with a rise in fixed assets by about 4.5 percentage points when carbon prices are high (text statement).
  - Unregulated (non-ETS) firms during high carbon price periods:
    - Small increases in input costs; turnover rises in lockstep with input costs, resulting in no profitability impacts.
- Investment composition (tangible vs intangible; Figure 5 and discussion):
  - ETS firms with payable emissions in periods of high carbon prices:
    - Commensurate increases in total fixed assets and tangible fixed assets, implying intangible fixed asset investments are also increasing.
  - ETS firms without payable emissions in periods of high carbon prices:
    - Only tangible fixed assets increase (7 percentage points); total fixed assets remain unchanged, implying a decline in investment in intangibles.
  - Interpretation: when firms must pay for emissions, they switch investment portfolio towards greater intangible capital, consistent with literature that greater intangible capital reduces energy intensity of production.
  - Supporting references in text: Yang and Shi (2018); IMF (2023) firm surveys in Germany; Wu and Wang (2022); De Jonghe, Mulier and Schepens (2020).

*Source: IMF Working Paper — section "3.  Data and Empirical Specification" (extracted from the supplied PDF content).*

### 5. Robustness

### 5. Robustness

### Robustness checks performed
- Controlled for additional firm-level characteristics to address omitted variable bias: included cash-to-asset ratio and leverage (debt-to-asset ratio) as proxies for firm resilience and amplification to external shocks.
- Verified sensitivity to an alternative measure of investment: tangible investment proxied by changes in tangible fixed assets.
- Explored heterogeneity by carbon price environment:
  - Periods of high carbon prices (dummy = 1 when carbon prices exceed the 75th percentile).
  - Periods of low carbon prices (dummy included in baseline regression).
  - Periods of high carbon price volatility (dummy = 1 when the standard deviation of carbon prices is above the 75th percentile).
- Estimated quantile regressions (median, q = 0.5) to check whether main effects are driven by distributional irregularities across firms.
- Sample and data: panel regressions for 12 European countries covering 1995–2020 based on EU ETS and Bureau van Dijk ORBIS data.

### Main robustness findings
- Inclusion of (lagged) cash-to-asset and debt-to-asset ratios: main results remain robust (see Annex Table 1).
- Estimated impacts of more stringent market-based climate policies on changes in input costs, turnover, profits, and fixed assets remain broadly the same as in the baseline across all ETS-regulated firm subgroups, including during periods of high carbon prices.
- Precision of some parameter estimates improves with additional controls.
- Difference from baseline: more stringent climate policies are associated with slightly larger increase in fixed assets (by 6 percentage points compared to 4.5 in the baseline).

### High vs low carbon price environments
- When carbon prices are low:
  - No change in input costs or profits of ETS firms (lack of excess profits from permit trading cited).
  - ETS firms with payable emissions: tightening of EPS when carbon prices are low is associated with almost 9 percent decline in input costs and a 7 percent decline in turnover (partly passed-through).
  - More stringent climate policies do not enhance investment by ETS firms with payable emissions in a low carbon price environment.
- When carbon prices are high:
  - ETS-firms with payable emissions increase investment levels during periods of high carbon prices (investment impact confirmed in both average and median analyses).

### Carbon price volatility
- During periods of high carbon price volatility:
  - No significant response of ETS firms with payable emissions to tightening of climate policies (Figure 7).
  - Higher volatility does not induce ETS firms to accumulate fixed assets.
  - ETS firms with no payable emissions reduce their fixed assets on average despite benefiting from volatile carbon prices.
  - Interpretation: regulated firms respond more to tighter budget constraints (e.g., payable emissions / high prices) than to carbon price uncertainty.

### Quantile (median) regression results
- Median effects (q = 0.5) are largely not statistically different from baseline results at the 5% significance level, indicating results are representative of at least 50% of ETS firms.
- Median effects for input costs and turnover become modestly negative among ETS firms with payable emissions in periods of high carbon prices — these are the only statistically distinct median results from baseline.
- Median profits of ETS firms remain protected or even increase in periods of high carbon prices.
- Investment impact among ETS firms with payable emissions: median effects roughly 3%, modestly weaker than average results but still statistically significant and economically meaningful.
- Conclusion: quantile results confirm robustness of main findings and suggest carbon price investment incentives are reflected in the broader ETS market, with some firms undertaking more ambitious investment.

### Implications highlighted in robustness section
- More stringent market-based climate policies do not negatively impact the average firm or ETS firms; they incentivize investment in low-carbon technologies and emission reductions.
- Cost pass-through is limited for most firms; only a few firms pass increased input costs downstream.
- Policy design and coordination matter to avoid fragmentation and to address distributional effects across firms (especially hard-to-abate sectors).
- Future research directions suggested: structural analysis of regulatory shock transmission, role of firm innovation capacity and distance to technological frontier in investment responses, and quantile regression analysis beyond the median to assess tail risks.

*Source: IMF Working Paper — Firms’ Response to Climate Regulations (section 5: Robustness; Annex Table 1).*

### Annex Table 2. Estimates of Tightening of Market-Based Policy Stringency: High vs Low Carbon Prices

### Annex Table 2. Estimates of Tightening of Market-Based Policy Stringency: High vs Low Carbon Prices

### Regression setup and sample
- Sample: 12 European countries covering 1995–2020; data cleaning follows Diez, Fan, and Villegas-Sánchez (2021).
- Data sources: EU Emissions Trading System; European Investment Bank; IMF, World Economic Outlook database; ORBIS; Organisation for Economic Co-operation and Development.
- Dependent variables: changes in input costs (material costs), turnover, profits (earnings before interest, tax, depreciation, and amortization), and fixed assets (in logarithms).
- Key regressors:
  - ∆EPS Market and ∆EPS Non-Market: changes in OECD market-based and non-market-based Environmental Policy Stringency indices.
  - ETS firms: firms with ETS-registered installations.
  - Payable emissions: verified emissions minus free allowances.
  - “Low carbon price” dummy = 1 in years with carbon prices below the 25th percentile.
  - “High carbon price” dummy = 1 in years with carbon prices above the 75th percentile.
- Estimation details: panel regression with firm, country-sector, and year-sector fixed effects; standard errors available from authors; ***, **, * denote statistical significance at 1 percent, 5 percent, and 10 percent levels, respectively. ETS = Emissions Trading System.

### Coefficient estimates (selected regressors and interactions)
- Lag ∆EPS Market:
  - -0.00989***, -0.00587***, -0.0131***, -0.0101***, 0.0104***, 0.0214***, 0.00296**, -0.00541***
- Lag ∆EPS Market × High carbon price:
  - 0.0503***, 0.0465***, 0.0537***, 0.0509***, 0.0160***, 0.00551, -0.000476, 0.00746**
- Lag ∆EPS Market × Low carbon price:
  - -0.0322***, -0.0240***, -0.0880***, 0.0670***
- Lag ∆EPS Market × ETS firm:
  - -0.0338***, -0.0435***, -0.0113*, -0.0149**, -0.0461**, -0.0648***, -0.0113, -0.00798
- Lag ∆EPS Market × ETS firm × High carbon price:
  - -0.0487**, -0.0382, -0.0293**, -0.0254*, 0.154***, 0.175***, -0.0163, -0.0208
- Lag ∆EPS Market × ETS firm × Low carbon price:
  - 0.0837**, 0.0267, 0.140*, -0.00146
- Lag ∆EPS Market × ETS firm with payable emissions:
  - 0.0102, 0.0503*, -0.0118, 0.00622, 0.0511, 0.0447, 0.0143, 0.0458**
- Lag ∆EPS Market × ETS firm with payable emissions × High carbon price:
  - 0.101**, 0.0602, 0.0435*, 0.0253, -0.228***, -0.232***, 0.0559*, 0.0305
- Lag ∆EPS Market × ETS firm with payable emissions × Low carbon price:
  - -0.140***, -0.0530**, -0.0826, -0.0725**
- Lag Log Employment:
  - -0.0538***, -0.0537***, -0.0523***, -0.0523***, 0.000802, 0.00105, 0.0184***, 0.0182***
- Lag Log Total Assets:
  - -0.130***, -0.130***, -0.112***, -0.112***, -0.245***, -0.245***, -0.187***, -0.187***
- Lag ∆EPS Non-Market:
  - 0.0109***, 0.0114***, 0.00169***, 0.00204***, -0.00408***, -0.00278***, -0.00490***, -0.00589***
- Lag Real GDP growth:
  - 0.0148***, 0.0145***, 0.0131***, 0.0129***, 0.0179***, 0.0170***, 0.00691***, 0.00765***
- Lag Output gap:
  - -0.00905***, -0.00901***, -0.00592***, -0.00590***, -0.0120***, -0.0119***, -0.000905***, -0.000983***
- Constant:
  - 1.890***, 1.889***, 1.651***, 1.650***, 3.354***, 3.353***, 2.569***, 2.570***

### Model fit, observations, and fixed effects
- Observations: 7,326,460; 7,326,460; 7,326,460; 7,326,460; 7,326,460; 7,326,460; 7,326,460; 7,326,460
- R-squared: 0.126, 0.126, 0.201, 0.201, 0.091, 0.091, 0.135, 0.135
- Fixed effects included:
  - Firm Fixed Effect: Yes
  - Country-Sector Fixed Effect: Yes
  - Year-Sector Fixed Effect: Yes

### Key patterns evident in the estimates
- Direct effect of Lag ∆EPS Market on firms (first set of coefficients) is generally negative and statistically significant across several specifications (e.g., -0.00989***, -0.0131***, -0.0101***).
- Interaction with High carbon price tends to offset or reverse the negative direct effect in many specifications (e.g., Lag ∆EPS Market × High carbon price: 0.0503***, 0.0537***, 0.0509***).
- ETS firm interactions show heterogeneous effects:
  - Lag ∆EPS Market × ETS firm often more negative (e.g., -0.0338***, -0.0435***, -0.0648***).
  - Lag ∆EPS Market × ETS firm × High carbon price shows mixed signs and magnitudes, including large positive coefficients in some columns (e.g., 0.154***, 0.175***).
- Payable emissions interact with market stringency to produce distinctive responses:
  - Lag ∆EPS Market × ETS firm with payable emissions × High carbon price includes large negative coefficients in some columns (e.g., -0.228***, -0.232***), and positive coefficients in others (e.g., 0.101**).
  - Lag ∆EPS Market × ETS firm with payable emissions × Low carbon price are negative and statistically significant in multiple cases (e.g., -0.140***, -0.0725**).

*Source: Annex Table 2, “Estimates of Tightening of Market-Based Policy Stringency: High vs Low Carbon Prices,” Firms’ Response to Climate Regulations: Empirical Investigations Based on the European Emissions Trading System, Working Paper No. WP/2024/135, International Monetary Fund.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024135-print-pdf.pdf_