## Environmental Policies and Innovation in Renewable Energy — 1. Introduction

## Source details

**Canonical URL:** [Environmental Policies and Innovation in Renewable Energy — 1. Introduction](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023180-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023180-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023180-print-pdf.pdf.json)

---

### Motivation and role of innovation
- Climate change requires drastic emission reductions and a green transition; innovation reduces renewable energy production costs and facilitates adoption.
- Green innovation and diversification enhance economic resilience to climate shocks.
- Innovation responds to localized competences, specialized human capital, firm adaptability, and government regulations and policies.
- Environmental regulations have ambiguous effects on innovation:
  - Negative channel: additional burdens on firms, weakening incentives to invest and innovate.
  - Positive channels: direct support (credit and subsidies to R&D), inducement to modernize production techniques, increased demand for clean energy raising expected returns on green innovation.

### Research question, scope, and data
- Objective: investigate the dynamic response of green innovation to climate change policies (CCPs).
- Sample and coverage:
  - 40 countries, 5 sectors, and 22 years → total of 4,400 observations.
  - Period: 2000-2021.
- Measures:
  - CCPs measured by the OECD Environmental Policy Stringency Index (EPS).
  - Green innovation measured by number of new patents related to green technologies from the IRENA (2022) database, classified by country, year, and sector (Industry, Building, Power, Transport, Waste).
- IRENA dataset highlights:
  - 140 thousand patents filed for renewable energy worldwide, classified by 6 economic sectors, for a sample of 64 economies during 2000-21.
  - Sample restricted to 40 countries with EPS information and to 5 sectors associated with NAICS codes.
  - Overall new patents in 2000-2021 grew five times, from about 50 to 250 thousand, with a sudden stop due to the COVID-19 crisis.
  - Prior to COVID-19, the power sector accounted for about a half of total new patents, followed by transport.
  - Top 10 innovator countries account for more than 90 percent of total new patents; “all other countries” share shrinks year-by-year.
  - China’s share increased from about 6 percent in 2000 to about 65 percent in 2019.
  - Japan’s share dropped to about 7 percent in 2019, declining 30 percentage points from 2000.
  - US and Korea shares remained in the ranges 15-20 percent and 6-10 percent, respectively.

### Data on Climate Change Policies (CCPs) and EPS
- EPS index details:
  - EPS values in the sample range from 0.83 in New Zealand to 4.89 in France.
  - EPS available for 40 countries during 1990-2020.
  - Average change of about 0.09, bounded between -.84 (minimum change) and 1.5 (maximum change) over the period.
  - Rapid increase in EPS since 2000 following regulatory waves and tightening of emissions regulations and R&D subsidies.
  - When EU ETS entered into force in 2005, the median change in EPS was about 0.47, which is 11.75 times the sample median.
  - Figure 3, panel A: average yearly change of EPS across countries ranges from approximately 0.03 in New Zealand to 1.6 in France.
  - Figure 3, panel B: distribution of average EPS across countries in 2020 shows substantial heterogeneity.
- OECD EPS disaggregated into:
  - market-based instruments,
  - non-market-based instruments,
  - technology-support instruments.

### Empirical strategy and methodology
- Four empirical steps:
  1. Macro-level dynamic response:
     - Local projection approach (Jordà, 2005) to estimate evolution of green patent applications after an increase in CCP stringency.
     - Finding: CCPs increase green patents; effects grow over time; positive effects primarily for non-market-based and technology-support policies.
  2. Address endogeneity:
     - Use an IV strategy exploiting cross-sectional variation in countries’ exposure to climate risks and time-varying global climate-related events.
  3. State-dependent responses:
     - Local projection smooth transition (Auerbach and Gorodnichenko, 2013) to allow heterogeneous responses by product market competition, economic uncertainty, financial stress, and demand conditions.
  4. Sectoral analysis:
     - Difference-in-differences approach (Rajan and Zingales, 1998) assuming weaker CCP effects for sectors with tighter financial constraints; comprehensive fixed effects including country-time fixed effects.

### Baseline econometric setup
- Equation (1) impulse-response baseline:
  - Dependent variable: percent variation in patenting activity in country i, sector s, between t+k and t, with k=1,...,5.
  - Regressor: ∆EPS_{i,t} (yearly change in EPS).
  - Controls include 2 lags of dependent variable and of ∆EPS; country-sector time trends; standard errors clustered at country/sector level.
- Horizons estimated: k=1,..,5.

### Baseline results — effects of EPS on green patents
- Impact of a 1 standard deviation increase in EPS (roughly a yearly change of EPS of 0.24 point):
  - Increases new green patents by about 4 percent one year after the policy change.
  - Increases new green patents by about 18 percent five years after the policy change.
- Large reforms (changes in EPS at the 99th percentile ≈ 0.91) imply:
  - An increase in green patenting of about 65 percent.
- Comparison to Zhang et al. (2022):
  - Zhang et al.: a 1-point increase in EPS increases green innovation by about 57 percent.
  - Translating this study’s results to a 1-point EPS increase yields:
    - short-term effect ≈ 16 percent,
    - medium-term (5-year) effect ≈ 80 percent.
- Robustness checks (qualitatively similar):
  - Additional controls: GDP growth, financial stress index, oil prices.
  - Change number of lags from 2 to 3 and 4.
  - Account for contemporaneous effects of EPS changes.
  - Exclude top and bottom 1 and 5 percent of dependent variable.
  - Exclude one country and one year at a time.
  - Control for lagged stock of patents at country level.
  - Annex Figures A2a–A2g report robustness results.

### Heterogeneity by policy instrument
- Results by CCP subcomponents:
  - Non-market-based and technology-support CCPs: positive and statistically significant effects on green patents.
  - Market-based policies: effect on green patents not statistically different from zero in these estimates.
- Note: market-based policies still contribute to emissions reductions and generate resources to compensate CCP costs; optimal policy-mix may include both types.

### Instrumental Variable (IV) analysis — addressing endogeneity
- IV strategy:
  - Instrument ∆EPS with interaction of a time-varying global term (number of flood events) and a country-specific term (length of coastline).
  - Alternative instruments tested: number of major hurricanes * minimum distance of country centroid to coast; number of people affected by earthquakes * share of urban population; number of wildfires around the globe per annum * agricultural land (km2) per capita.
- First-stage diagnostics:
  - Kleibergen–Paap rk Wald F statistic ranges from 85.9 (for t=4) to 97.2 (for t=5).
  - These are approximately 7 times the Stock-Yogo critical value for strong instruments (16.38).
- First-stage coefficients (Flood_events*coastal_length):
  - t=0: .00007*** (.00000)
  - t=1: .00008*** (.00000)
  - t=2: .00008*** (.00000)
  - t=3: .00008*** (.00000)
  - t=4: .00008*** (.00000)
  - t=5: .00008*** (.00000)
  - Observations by horizon: 2664, 2664, 2664, 2646, 2599, 2418 respectively.
- IV second-stage results:
  - Effect of a 1 standard deviation increase in EPS on green innovation is larger with IV, consistent with OLS baseline estimates being biased toward zero.

### State-dependent effects (smooth transition framework)
- State variables (normalized to zero mean and unit variance):
  - Product market regulation (competition indicator ranging from -1 to 1; higher = more liberalization/more competition).
  - Business cycle: GDP growth.
  - Uncertainty: World Uncertainty Index (WUI).
  - Financial stress: Romer and Romer (2017) discrete measure.
- Smooth transition parameter γ set to 5 (results unchanged for γ=2.5 or γ=7).
- Key state-dependent findings:
  - GDP growth (expansions vs recessions):
    - Positive CCP effects on green patents are larger during expansions — about 1.5 times baseline magnitude.
    - Difference between low and high growth regimes statistically significant for most horizons (F-tests: t=0 14.60***; t=3 3.30*; t=4 9.95***; others reported).
  - Uncertainty:
    - Environmental policy stimulates green innovation more when uncertainty is low.
    - Difference statistically significant across all horizons (F-tests: t=0 10.06***; t=1 14.35***; t=2 3.41*; t=3 11.89***; t=4 9.39***; t=5 5.36**).
  - Financial stress:
    - When financial stress is high, CCP impact is not statistically significant.
    - When financial stress is low, impact is large and precisely estimated; difference statistically significant in medium term (F-tests: t=4 2.79*; t=5 4.03**).
  - Product market regulation (competition):
    - Effect of CCPs on green patents is larger when competition is high.
    - Difference between low and high competition regimes highly statistically significant across all horizons (F-tests: t=0 3.65*; t=1 11.29***; t=2 8.90***; t=3 10.44***; t=4 7.51***; t=5 17.48***).

### Sectoral analysis — external financial dependence heterogeneity
- Identification:
  - Difference-in-differences (Rajan and Zingales, 1998) with EFD = (total capital expenditures − current cash flow) / total capital expenditures using US Compustat firm-level data aggregated to sector median, matched via NAICS.
  - Fixed effects: country-sector, country-time, sector-time; lags l=0,1,2 included.
  - EFD used as ranking (1..5) and as continuous median score.
- Sector ranking by EFD (Table A4):
  - Rank 1: Power
  - Rank 2: Transport
  - Rank 3: Waste
  - Rank 4: Industry
  - Rank 5: Building
- Findings:
  - CCP effects on green patenting are higher for sectors with low financial constraints.
  - Quantified differential (25th vs 75th percentile of EFD):
    - Short term (one year): about 1.5 percentage points higher green patent growth for low-EFD industry.
    - Medium term (five years): about 4 percentage points higher green patent growth for low-EFD industry.

### Descriptive statistics (selected)
- Patent: Obs. 4782 Mean 544.859 Std. Dev. 3652.181 Min 0 Max 100429 (IRENA).
- Patent (log): Obs. 4782 Mean 2.887 Std. Dev. 2.372 Min 0 Max 11.517 (IRENA).
- CCP: Obs. 4782 Mean 2.3 Std. Dev. 1.101 Min 0 Max 4.889 (OECD).
- ∆CCP: Obs. 4782 Mean .093 Std. Dev. .249 Min -.833 Max 1.5 (OECD).
- CCP_mkt: Obs. 4782 Mean 1.183 Std. Dev. .861 Min 0 Max 4.167 (OECD).
- CCP_non_mkt: Obs. 4782 Mean 3.943 Std. Dev. 1.689 Min 0 Max 6 (OECD).
- Gdp_growth: Obs. 4782 Mean 2.476 Std. Dev. 3.37 Min -14.629 Max 25.176 (OECD).
- WUI: Obs. 4476 Mean .214 Std. Dev. .167 Min 0 Max 1.343 (Ahir et al., 2022).
- PMR index: Obs. 3324 Mean .186 Std. Dev. .398 Min -1 Max 1 (Alesina et al., 2023).
- EFD: Obs. 3489 Mean -.430 Std. Dev. .463 Min -.961 Max .232 (Compustat).

### Main contributions and policy-relevant implications
- Dynamic setting: analyzes short- and medium-term responses, accounting for patenting recognition lags.
- Heterogeneity and timing: responses larger in countries with greater product market competition and magnified during stronger economic activity with low uncertainty and low financial stress — highlighting importance of timing and complementary policies.
- Improved identification: IV and sectoral difference-in-differences with 3-dimensional fixed effects strengthen causal claims.
- Policy design implications:
  - Emphasize non-market-based policies (for example, R&D subsidies) to spur green innovation.
  - Implement CCPs when the economic environment is stronger and financial stress and uncertainty are lower to maximize innovation response.
  - Ensure pro-competitive regulatory frameworks to amplify innovation responses.

*Source: wpiea2023180-print-pdf*

### 1. Introduction

### 1. Introduction

### Motivation and role of innovation
- The fight against climate change requires countries to drastically reduce emissions and facilitate the green transition; innovation plays a key role by reducing the cost of renewable energy production and facilitating adoption of green energy.
- Green innovation and diversification of energy sources can enhance economic resilience to climate shocks.
- Innovation is money- and time-expensive and responds to drivers such as localized competences, specialized human capital, firm adaptability, and government regulations and policies.
- Environmental regulations have ambiguous effects on innovation:
  - Potential negative channel: impose additional burdens on firms and weaken incentives to invest and innovate.
  - Potential positive channels: direct support (credit and subsidies to R&D), inducement to modernize production techniques, and increased demand for clean energy raising expected returns on green innovation.

### Research question, scope, and data
- Objective: investigate the dynamic response of green innovation to climate change policies (CCPs).
- Sample and coverage:
  - 40 countries, 5 sectors, and 22 years → total of 4,400 observations.
  - Period: 2000-2021.
- Measures:
  - CCPs measured by the OECD Environmental Policy Stringency Index (EPS).
  - Green innovation measured by the number of new patents related to green technologies from the IRENA (2022) database, classified by country, year, and sector of application (Industry, Building, Power, Transport, Waste).
- IRENA dataset details:
  - 140 thousand patents filed for renewable energy worldwide, classified by 6 economic sectors, for a sample of 64 economies during 2000-21.
  - Sample restricted to 40 countries with EPS information and to 5 sectors associated with NAICS codes (Industry, Transport, Building, Waste, Power).
  - Overall number of new patents in 2000-2021 grew by five times, from about 50 to 250 thousand, experiencing a sudden stop due to the COVID-19 crisis.
  - Prior to COVID-19, the power sector accounted for about a half of total new patents, followed by the transport sector.
  - Top 10 innovator countries account for more than 90 percent of the total number of new patents; the share of “all other countries” shrinks year-by-year.
  - China’s share increased from about 6 percent in 2000 to about 65 percent in 2019.
  - Japan’s share dropped to about 7 percent in 2019, declining 30 percentage points from 2000.
  - US and Korea shares remained in the ranges 15-20 percent and 6-10 percent, respectively.

### Empirical strategy and methodology
- Four main empirical steps:
  1. Macro-level dynamic response:
     - Use the local projection approach (Jordà, 2005) to estimate evolution of green patent applications following an increase in CCP stringency.
     - Finding: CCPs increase green patents, with effects gradually increasing over time; statistically significant positive effects primarily for non-market-based policies (e.g., emission limits and R&D subsidies) and technology-support policies.
  2. Address endogeneity:
     - Potential reverse causality (countries more prone to implement CCPs when green innovation is weak) could bias OLS estimates toward zero.
     - Use an instrumental variable (IV) strategy following Furceri et al. (2022) exploiting cross-sectional variation in countries’ exposure to climate risks and time-varying global climate-related events.
  3. State-dependent responses:
     - Allow response of green innovation to CCPs to vary across countries and economic conditions using a local projection smooth transition approach (Auerbach and Gorodnichenko, 2013).
     - Examine mediation by product market competition, economic uncertainty, financial stress, and demand conditions.
  4. Sectoral analysis:
     - Use a difference-in-differences approach (Rajan and Zingales, 1998) under the assumption that CCPs have weaker effects for sectors with tighter financial constraints.
     - Include a comprehensive set of fixed effects (including country-time fixed effects) to control for unobserved cross-country heterogeneity in macroeconomic conditions and strengthen causal identification.

### Key mediating factors and hypotheses
- Uncertainty:
  - Expect larger CCP effects on green innovation during periods of low uncertainty.
  - Use the World Uncertainty Index by Ahir et al. (2022).
- Financial system health:
  - Access to finance is a key barrier to innovation; financial distress reduces loan supply and may negatively affect green innovation.
  - Use Romer and Romer (2017) measure of financial stress to proxy financial system health.
- Business cycle:
  - Innovation is expected to be procyclical; CCP effects should be larger during economic expansions.
- Product market competition:
  - Competition can either hinder or spur innovation; recent empirical studies largely find a positive effect of competition on investment and innovation.
  - Use an indicator of product market regulation (Alesina et al., 2023) to assess mediation—expect higher pro-competition reforms to amplify CCP effects.

### Main contributions
- Dynamic setting: analyzing short- and medium-term responses of green innovation to CCPs accounts for temporal lag between policy adoption and innovation output (patenting recognition lag).
- Heterogeneity and timing: show that response of green innovation to CCPs is larger in countries with greater product market competition and is magnified during stronger economic activity—highlighting timing and complementary policies importance.
- Improved identification: strengthen causal claims using IV and sectoral difference-in-differences approaches, and a 3-dimensional fixed effects setup.

### Data on Climate Change Policies (CCPs) and EPS
- EPS index:
  - OECD Environmental Policy Stringency Index measures degree of stringency of environmental regulation (higher values = more stringent).
  - EPS available for 40 countries during 1990-2020.
  - Distributional statistics and dynamics noted:
    - Average change of about 0.09, bounded between -.84 (minimum change over the period) and 1.5 (maximum change over the period).
    - Rapid increase in EPS since 2000 following regulatory waves and tightening of emissions regulations and R&D subsidies.
    - When the EU Emissions Trading System (EU ETS) entered into force in 2005, the median change in EPS index was about 0.47, which is 11.75 times the sample median.
    - Figure 3, panel A: average yearly change of EPS across countries ranges from approximately 0.03 in New Zealand to 1.6 in France.
    - Figure 3, panel B: distribution of average EPS across countries in 2020 shows substantial heterogeneity.

### Organization of the paper (from the original chapter)
- Section 2: overview of the literature.
- Section 3: data used in the empirical analysis.
- Section 4: macro-level response of green innovation.
- Section 5: sectoral difference-in-differences analysis.
- Section 6: conclusions summarizing main results and policy implications.

*Environmental Policies and Innovation in Renewable Energy — 1. Introduction (wpiea2023180-print-pdf)*

### 0.83 in New Zealand to 4.89 in France.

### Environmental Policies and Innovation in Renewable Energy

### OECD EPS composition and measurement
- EPS values in the sample range from 0.83 in New Zealand to 4.89 in France.
- The OECD database provides disaggregated climate stringency indices classified as:
  - market-based instruments (such as taxes on emissions),
  - non-market-based instruments (such as emission limits),
  - technology-support instruments (such as low-carbon R&D expenditures).
- The empirical analysis distinguishes these subcomponents to understand dynamic responses of green innovation to CCPs; Appendix Figure A1 shows country breakdowns by EPS sub-component.

### Macro-level analysis — baseline estimates (equation and dataset)
- Estimation approach:
  - Impulse-response functions following Jordà (2005); regression eq. (1) estimates percent change in renewable energy patents between t and t+k.
  - Country-sector specific time trends: country fixed effects * sector fixed effects * a time trend.
  - ∆EPSi,t measures yearly variation in environmental policy stringency.
  - Specification includes lags l=0,1,2 of dependent variable and ∆EPS to account for serial correlation.
- Sample and estimation:
  - Balanced panel of 40 countries, across 5 sectors, over the period 2000-2021.
  - Horizons k=1,..,5 (estimated for each horizon).
  - Robust standard errors clustered at the country/sector level (results robust to clustering at the country level).

### Baseline results — effects of EPS on green patents
- Impact of a 1 standard deviation increase in EPS (roughly corresponding to a yearly change of EPS of 0.24 point):
  - Increases number of new green patents by about 4 percent one year after the policy change.
  - Increases number of new green patents by about 18 percent five years after the policy change.
- Large reforms (changes in EPS at the 99th percentile of sample distribution ≈ 0.91) imply:
  - An increase in green patenting of about 65 percent.
- Comparison to related findings:
  - Zhang et al. (2022): a 1-point increase in EPS increases green innovation by about 57 percent.
  - Translating this study’s results to a 1-point EPS increase yields:
    - short-term effect ≈ 16 percent,
    - medium-term (5-year) effect ≈ 80 percent.
- Robustness checks performed (results qualitatively similar):
  - Include additional controls: GDP growth, financial stress index, oil prices.
  - Change number of lags from 2 to 3 and 4.
  - Account for contemporaneous effects of EPS changes.
  - Exclude top and bottom 1 and 5 percent of dependent variable.
  - Exclude one country and one year at a time.
  - Control for lagged stock of patents at country level.
  - Annex Figures A2a–A2g report robustness results.

### Heterogeneity by policy instrument
- Estimating eq. (1) using market-based, non-market-based, and technology-support CCP measures:
  - Non-market-based and technology-support CCPs: positive and statistically significant effects on green patents.
  - Market-based policies: effect on green patents not statistically different from zero in these estimates.
- Note: Market-based policies still contribute to emissions reductions and generate resources to compensate CCP costs; optimal policy-mix may include market- and non-market-based instruments.

### Instrumental Variable (IV) analysis — addressing endogeneity
- Endogeneity concerns:
  - Reverse causality: low green innovation could induce more stringent policies.
  - Measurement error in policy reform indicators.
- IV strategy:
  - Instrument ∆EPS with interaction of a time-varying global term and a constant country-specific term.
  - Global term: number of flood events (environmental pressure for policy action).
  - Country term: geographical characteristic — length of the coastline.
  - Alternative instruments tested (qualitatively similar): number of major hurricanes * minimum distance of country centroid to coast; number of people affected by earthquakes * share of urban population; number of wildfires around the globe per annum * agricultural land (km2) per capita.
- First-stage diagnostics:
  - Kleibergen–Paap rk Wald F statistic ranges from 85.9 (for t=4) to 97.2 (for t=5).
  - These values are approximately 7 times the Stock-Yogo critical value for strong instruments (16.38).
- IV second-stage results:
  - Effect of a 1 standard deviation increase in EPS on green innovation is larger with the IV approach, consistent with OLS baseline estimates being biased towards zero.

### State-dependent effects (smooth transition framework, eq. (3))
- State variables considered (z, normalized to zero mean and unit variance):
  - Product market regulation (competition indicator ranging from -1 to 1; higher = more liberalization/more competition).
  - Business cycle: GDP growth (within-country variation exploited).
  - Uncertainty: World Uncertainty Index (WUI).
  - Financial stress: Romer and Romer (2017) discrete measure.
- Smooth transition function F(z) used with γ set to 5 (results unchanged for γ=2.5 or γ=7).
- Key state-dependent findings (Figures 8–11; Table 2 F-tests for regime differences):
  - GDP growth (expansions vs recessions):
    - Positive effects of CCPs on green patents are larger during economic expansions — about 1.5 times baseline magnitude.
    - Difference between low and high growth regimes statistically significant for most horizons.
  - Uncertainty:
    - Environmental policy stimulates green innovation more when uncertainty is low.
    - Difference between low and high uncertainty regimes statistically significant across all horizons.
  - Financial stress:
    - When financial stress is high, impact of CCPs is not statistically significant.
    - When financial stress is low, impact is large and precisely estimated; difference statistically significant in medium term.
  - Product market regulation (competition):
    - Effect of CCPs on green patents is larger when competition is high.
    - Difference between low and high competition regimes highly statistically significant across all horizons.

### Sectoral analysis — external financial dependence (EFD) heterogeneity (eq. (4))
- Identification:
  - Difference-in-differences approach following Rajan and Zingales (1998).
  - EFD measure: ratio of total capital expenditures minus current cash flow to total capital expenditures; constructed using US Compustat firm-level data aggregated to sector-level (median), matched via NAICS codes.
  - Specification includes country-sector, country-time, and sector-time fixed effects; lags l=0,1,2 included.
  - EFD used alternatively as ranking (1..5) and as continuous median score; Table A4 reports sector ranking.
- Findings (Figure 12 and Appendix Figures A3):
  - Effects of CCPs on green patenting are higher for sectors with low financial constraints.
  - Quantified differential:
    - For an industry at the 25th percentile of EFD (low external financial dependence) versus 75th percentile (high EFD):
      - Short term (one year after policy change): about 1.5 percentage points higher green patent growth for low-EFD industry.
      - Medium term (five years after): about 4 percentage points higher green patent growth for low-EFD industry.

*Source: IMF Working Paper — Environmental Policies and Innovation in Renewable Energy (excerpts, wpiea2023180-print-pdf).*

### 6. Conclusions

### 6. Conclusions

### Context and motivation
- Climate change is (one of) the greatest challenge of our time.
- The use of conventional energy is the principal cause of global warming and climate change, leading to a series of issues for the society, such as natural disasters and weather extreme events.
- The transition to green energy is thus becoming key to ensure the sustainability of the planet.
- To stimulate the reduction of greenhouse emissions and ease the spread of renewable energy, most governments attempt to formulate and implement numerous environmental policies.
- The effect of CCPs on national economies may be ambiguous, as noted by several studies (see OECD, 2021, for a review): on the one side, CCPs may negatively affect the economy by imposing additional costs on firms; on the other side, they may stimulate the willingness of firms to invest and innovate (Porter, 1996).

### Research approach
- The article offers a dynamic analysis of the extent to which CCPs affects the production of green innovation.
- Measures used:
  - Environmental Policy Stringency (EPS) index, provided by the OECD, to measure the degree of environmental policies stringency.
  - Data on new patents filed for renewable energy to proxy green innovation.

### Key findings
- The production of green innovation drastically increases when CCPs become more stringent.
- A 1-standard deviation increase in EPS positively fosters green patent activity by about the 18 percent, five years after the policy shock.
- Major reforms provide illustrative magnitude: the introduction of the EU Emissions Trading System (ETS) in 2005 increases green patenting by about the 69 percent in the medium term.

### Heterogeneity and policy-relevant modifiers
- Not all CCPs spur green innovation: the positive effects of CCPs are mostly related to non-market-based policies (such as R&D subsidies).
- The state of the economy at the time of CCPs implementation matters:
  - Effects are particularly strong in countries with more pro-competitive regulation.
  - Effects are particularly strong when the economic environment is strong and characterized by low uncertainty and financial stress.

### Policy design implications (implicit from findings)
- To maximize positive effects of CCPs on green innovation, policymakers should consider:
  - Emphasizing non-market-based policies (for example, R&D subsidies).
  - Implementing CCPs when the economic environment is stronger and financial stress and uncertainty are lower.
  - Ensuring pro-competitive regulatory frameworks to amplify innovation responses.

*Source: wpiea2023180-print-pdf - 6. Conclusions*

### References

### wpiea2023180-print-pdf - References

### Theoretical foundations and conceptual papers
- Key theoretical works cited:
  - Acemoglu, Aghion, Bursztyn, Hemous (2012). The Environment and Directed Technical Change. American Economic Review, 102(1), 131-166.
  - Dixit A.K., Pindyck R.S. (1994), Investment Under Uncertainty, Princeton University Press.
  - Schumpeter, J. A. (1942). Capitalism, Socialism and Democracy. Harper and Row, New York.
  - Porter, M. E. (1996). What is strategy? Harvard Business Review, 74(6), 61-78.
  - Porter, M. E., & Van der Linde, C. (1995). Toward a new conception of the environment-competitiveness relationship. Journal of economic perspectives, 9(4), 97-118.

### Empirical evidence on environmental policy, innovation, and renewable energy
- Representative empirical studies and findings included in the references:
  - Albrizio, Kozluk, Zipperer (2017). Environmental policies and productivity growth: Evidence across industries and firms. Journal of Environmental Economics and Management, 81: 209-226.
  - Dechezleprêtre, Sato (2017). The impacts of environmental regulations on competitiveness. Review of Environmental Economics and Policy, 11(2), 183-206.
  - Johnstone, Haščič, Popp (2010). Renewable energy policies and technological innovation: evidence based on patent counts. Environmental and resource economics, 45(1), 133-155.
  - Nesta, Vona, Nicolli (2014). Environmental policies, competition, and innovation in renewable energy. Journal of Environmental Economics and Management, 67(3), 396-411.
  - Hille, Althammer, Diederich (2020). Environmental regulation and innovation in renewable energy technologies: does the policy instrument matter? Technol. Forecast. Soc. Chang. 153, 119921.
  - Ren, Chen, Zhang (2018). Environmental regulation and green innovation in China: Evidence from patent data. Technological Forecasting and Social Change, 135, 221-229.
  - Zhang, Zheng, Feng, Chang (2022). Does an environmental policy bring to green innovation in renewable energy? Renewable Energy, 195: 1113-1124.

### Data sources, indices, and measurement
- Indices and datasets used or referenced:
  - EPS index (OECD data) — used throughout figures and analysis (e.g., Figures 2 and 3).
  - IRENA patent data (patent counts used; see Table A3: Patent Obs. 4782 Mean 544.859 Std. Dev. 3652.181 Min 0 Max 100429).
  - World Uncertainty Index (Ahir, Bloom, Furceri (2022)) — used in uncertainty analysis (WUI Obs. 4476 Mean .214 Std. Dev. .167 Min 0 Max 1.343; Table A3).
  - Romer & Romer (2017) financial distress index — used as financial stress (Financial stress Obs. 1830 Mean 1.428 Std. Dev. 2.401 Min 0 Max 11.5; Table A3).
  - BP Statistical Review of World Energy — oil price (Oil Price Obs. 4782 Mean 63.818 Std. Dev. 28.233 Min 24.444 Max 111.67; Table A3).
  - Compustat — External Financial Dependence (EFD) and Intangibility (EFD Obs. 3489 Mean -.430 Std. Dev. .463 Min -.961 Max .232; Intangibility Obs. 3489 Mean .0533 Std. Dev. .0421 Min 0 Max .112; Table A3).
  - PMR index (Alesina et al., 2023) — competition proxy (PMR index Obs. 3324 Mean .186 Std. Dev. .398 Min -1 Max 1; Table A3).
  - IRENA (2022), Renewable Technology Innovation Indicators; ISBN: 978-92-9260-424-0.

### Econometric specifications and identification strategies (as reported in figures and tables)
- Main empirical equations referenced (preserved notation as in source):
  - Equation (1) impulse-response baseline:
    - Dependent variable: percent variation in patenting activity in country i, sector s, between t+k and t, with k=1,...,5.
    - Regressor of interest: ∆CCP_{i,t} (yearly change in the EPS index in country i, between t and t-1).
    - Controls include 2 lags of the dependent variable and of the CCP shock. Standard errors clustered at country/sector level.
  - Equation (2) IV approach:
    - Uses predicted CCP shock ∆CCP̂_{i,t} with instrument = number of floods at global level at time t multiplied by length of coastline in country i.
  - Equation (3) smooth transition (regime-dependent effects):
    - Incorporates F(z_{it}) (smooth transition function) to separate low (L) and high (H) regimes (e.g., recession vs growth, low vs high uncertainty, low vs high financial stress, low vs high competition).
  - Equation (4) sectoral analysis with interaction:
    - Includes ∆CCP_{i,t} * EFD_s interaction; three batteries of fixed effects: country-year, country-sector, and sector-year.
- Instrumental variable first-stage results (Table 1, First stage):
  - Flood_events*coastal_lenght coefficients by horizon:
    - t=0: .00007*** (.00000)
    - t=1: .00008*** (.00000)
    - t=2: .00008*** (.00000)
    - t=3: .00008*** (.00000)
    - t=4: .00008*** (.00000)
    - t=5: .00008*** (.00000)
  - Observations: 2664 (t=0), 2664 (t=1), 2664 (t=2), 2646 (t=3), 2599 (t=4), 2418 (t=5).
  - KleibergenPaap_rk_Wald_F_statistic: 96.2, 95.0, 91.2, 92.6, 85.9, 97.2 respectively.
  - Stock-Yogo weak ID test critical value for 10% maximal IV size: 16.38.
  - Note: Standard errors clustered at country/sector level. Significance: *** p<0.01, ** p<0.05, * p<0.1.

- F-test differences between regimes (Table 2):
  - GDP growth: t=0 (14.60***), t=1 (0.03), t=2 (0.00), t=3 (3.30*), t=4 (9.95***), t=5 (1.75).
  - Uncertainty: t=0 (10.06***), t=1 (14.35***), t=2 (3.41*), t=3 (11.89***), t=4 (9.39***), t=5 (5.36**).
  - Financial stress: t=0 (0.10), t=1 (0.04), t=2 (0.34), t=3 (0.31), t=4 (2.79*), t=5 (4.03**).
  - Competition: t=0 (3.65*), t=1 (11.29***), t=2 (8.90***), t=3 (10.44***), t=4 (7.51***), t=5 (17.48***).
  - Note: *** p<0.01, ** p<0.05, * p<0.1.

### Figures, robustness checks, and appendices (inventory of visual/material content)
- Figures include (preserved captions and notes):
  - Figure 1. Evolution of patents by country (top 10 countries share, period 2000-2021).
  - Figure 2. Evolution of the EPS index over time (median, 25th percentile, 75th percentile) — authors elaboration on OECD data; x-axis years 2000 to 2020.
  - Figure 4–Figure 12. Impulse response functions showing impact of CCPs on green innovation under various specifications, robustness checks, decompositions (market-based, non-market-based, technology-support CCPs), IV approach, interactions with recession/growth, uncertainty, financial constraints, competition, and sectoral external finance dependence.
  - Appendix Figures A1–A3 and robustness panels A2a–A2h presenting additional checks: alternative lag structures (3 and 4 lags), contemporaneous CCP shock, exclusion of outliers (top/bottom 1% and 5%), leave-one-out country/year checks, control for lagged stock of patents, and sectoral EFD interactions.

### Sectoral and country coverage
- Economic sectors used (Table A1):
  - Building: NAICS 23
  - Industry: NAICS 31-33
  - Power: NAICS 22
  - Transport: NAICS 48
  - Waste: NAICS 562
- List of countries included in the analysis (Table A2): Australia, Austria, Belgium, Brazil, Canada, Chile, China, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, India, Indonesia, Ireland, Israel, Italy, Japan, Korea, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Russia, Slovak Republic, Slovenia, South Africa, Spain, Sweden, Switzerland, Turkey, United Kingdom, United States.
- Ranking of sectors by External Financial Dependence (Table A4):
  - Power N 4134 Rank 1
  - Transport N 4134 Rank 2
  - Waste N 4134 Rank 3
  - Industry N 4134 Rank 4
  - Building N 4134 Rank 5

### Descriptive statistics (Table A3 highlights)
- Patent: Obs. 4782 Mean 544.859 Std. Dev. 3652.181 Min 0 Max 100429 (Source IRENA).
- Patent (log): Obs. 4782 Mean 2.887 Std. Dev. 2.372 Min 0 Max 11.517 (IRENA).
- CCP: Obs. 4782 Mean 2.3 Std. Dev. 1.101 Min 0 Max 4.889 (OECD).
- ∆CCP: Obs. 4782 Mean .093 Std. Dev. .249 Min -.833 Max 1.5 (OECD).
- CCP_mkt: Obs. 4782 Mean 1.183 Std. Dev. .861 Min 0 Max 4.167 (OECD).
- CCP_non_mkt: Obs. 4782 Mean 3.943 Std. Dev. 1.689 Min 0 Max 6 (OECD).
- Gdp_growth: Obs. 4782 Mean 2.476 Std. Dev. 3.37 Min -14.629 Max 25.176 (OECD).
- WUI: Obs. 4476 Mean .214 Std. Dev. .167 Min 0 Max 1.343 (Ahir et al., 2022).
- PMR index: Obs. 3324 Mean .186 Std. Dev. .398 Min -1 Max 1 (Alesina et al., 2023).
- EFD: Obs. 3489 Mean -.430 Std. Dev. .463 Min -.961 Max .232 (Compustat).
- Intangibility: Obs. 3489 Mean .0533 Std. Dev. .0421 Min 0 Max .112 (Compustat).

*Italic: Source — wpiea2023180-print-pdf - References*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023180-print-pdf.pdf_