## Climate Policies as a Catalyst for Green FDI — Working Paper No. WP/2024/046 (selected empirical results)

## Source details

**Canonical URL:** [Climate Policies as a Catalyst for Green FDI — Working Paper No. WP/2024/046 (selected empirical results)](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024046-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024046-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024046-print-pdf.pdf.json)

---

### Data and classification
- Greenfield FDI data:
  - Source: fDi Markets database; coverage starts in 2003 and updated monthly.
  - Includes new projects and expansions; excludes mergers and acquisitions; joint ventures included if they lead to a new physical operation and majority foreign ownership.
  - Author creates a “green” label using clusters and tags (see Annex A); all other projects classified as non-green.
  - Sample restriction: countries with at least one project (of any type) in more than 15 years.
  - Pattern: green FDI accounted for 10 percent of total greenfield FDI between 2014 and 2017 and 40 percent of total investment by 2022.
  - Appendix A: about 13,000 projects labeled green over 20 years, out of close to 300,000 (about 4 percent).
- Aggregate FDI data:
  - Source: Financial Flows and Analytics (FFA) database constructed by the IMF’s Research Department; information for 165 countries dating back to 1970.
- Climate policies:
  - Main source: Climate Policy Database (CPD); selection includes policies with climate change mitigation as an objective (roughly 93 percent of policies).
  - Policies classified by budget impact into four categories: (i) revenue, (ii) expenses, (iii) regulations with no budget impact, (iv) nonregulatory budget neutral policies.
  - CPD limitation: does not contain stringency information; OECD’s EPS used as robustness.
- Country-level and bilateral variables: Penn World Tables v10.1, WDI, UNCTAD TRAINS tariffs for LCT goods, CEPII gravity database version 202211.

### Econometric approach (overview)
- Levels of aggregation: aggregate country-year, sectoral (ISIC one-digit grouping), and bilateral (gravity-style) analyses.
- Aggregate (country-year) specifications:
  - PPML for counts/flows with baseline: y_h,i,t = exp{α_i + β log(CP_i,t−1) + γ X_i,t−1} + ε_i,t, where h ∈ {total, green, non−green}.
  - Also estimate FDI_h,i,t / GDP_i,t = α_i + β log(CP_i,t−1) + γ X_i,t−1 + ε_i,t.
- Sectoral specification allows sector-specific elasticities φ_s via interactions and includes sector-time fixed effects σ_s,t and country trends.
- Bilateral (gravity) PPML specification:
  - x_h,i,j,t = exp{α_i + ω_j,t + β log(CP_i,t−1) + γ X_i,t−1 + δ Z_i,j,t−1} + ε_i,t.
  - Policy instrument heterogeneity modeled by shares of policy types share_i,p,t−1 and coefficients π_p; elasticity formulas provided (see equation 6).

### Main empirical findings — aggregate FDI (section 3.1)
- Core result:
  - A higher number of active climate policies is associated with larger green FDI inflows and projects.
- Statistical evidence (selected coefficients and significance):
  - Table 1:
    - log Number of climate policies (t-1) — Projects (column (1)): 0.6358*** (standard error: 0.1146)
    - log Number of climate policies (t-1) — Inflows (column (5)): 0.6341*** (standard error: 0.1400)
  - Table 2 (non-green):
    - log Number of climate policies (t-1) — Projects (column (1)): 0.1682*** (standard error: 0.0351)
    - log Number of climate policies (t-1) — Inflows (column (6)): 0.0971* (standard error: 0.0525)
  - Table 3 (total):
    - log Number of climate policies (t-1) — Projects (column (1)): 0.1790*** (standard error: 0.0361)
    - log Number of climate policies (t-1) — Inflows (column (6)): 0.1370*** (standard error: 0.0526)
- Economic magnitude:
  - A one standard deviation increase in the number of climate policies is associated with a 15 percent increase in green FDI inflows and projects.
  - Conservative estimate: increase in climate policy count from the EMDE median in 2019 to the 75th percentile relates to a 20 percent increase in green FDI flows/projects (using Tables 1; columns (3) and (7)).
- Robustness and endogeneity:
  - Country-specific linear trends, lagged dependent variables, and sub-sample exclusions leave main estimates materially unchanged.
  - IV approach (distance-weighted stock of policies abroad as instrument) yields larger IV coefficients than OLS:
    - Table 6 OLS (column (1)): log Number of climate policies (t-1) = 0.1628** (standard error: 0.0821)
    - Table 6 IV (column (2)): log Number of climate policies (t-1) = 0.2222** (standard error: 0.1058)
    - Table 6 IV (column (4)): log Number of climate policies (t-1) = 0.4376** (standard error: 0.2224)
  - Kleibergen-Paap Wald statistic above Cragg-Donald critical values (weak instrument null rejected).

### Sectoral heterogeneity (aggregate and bilateral)
- Aggregate sectoral interactions (Table 7, selected):
  - log Number of climate policies (t-1) — Green Projects (column (1)): 0.2183** (standard error: 0.0910)
  - Interaction — industry × log Number of climate policies — Green Projects (column (5)): 0.2616*** (standard error: 0.0929)
  - Interaction — energy × log Number of climate policies — Non-green Projects (column (3)): -0.2070*** (standard error: 0.0339)
- Interpretation:
  - Industry: complementarities between green and non-green FDI (both increase with climate policies).
  - Energy: substitution pattern—stronger climate agendas associated with higher green projects at the expense of non-green ones.
  - Other sectors: mixed or insignificant coefficients; adverse effects on non-green FDI concentrated in agriculture and energy, and lower non-green construction projects.

### Bilateral flows and identification (section 3.2)
- Estimation controls:
  - Destination fixed effects, source country-year fixed effects, trade agreement dummy, gravity controls (distance, language) or country-pair fixed effects; controls for destination capital stock and trade-weighted LCT and overall applied tariffs.
- Destination-country climate policies:
  - Larger number of destination climate policies associated with higher bilateral green FDI flows and green projects (Table 8):
    - Number of climate policies (in logs), destination country — Green Flows (column (1)): 0.2852* (standard error: 0.1548)
  - Magnitude: one standard deviation increase in recipient country policy stock associated with increase in bilateral green projects of 6 percent (or 7 percent on average for flows/projects as reported).
  - Results robust to country-pair dummies.
  - No systematic evidence that destination policies hinder non-green bilateral flows; negative link with bilateral non-green projects but smaller than green coefficients.
- Trade openness and LCT tariffs:
  - Trade-weighted LCT tariffs, destination country — negative and significant (Table 8, e.g., column (1): -0.1091***, standard error: 0.0160).
  - Interpretation: lower LCT tariffs reduce technology costs and support LCT diffusion via FDI; tariff-jumping motive less dominant than input-cost channel.
- Heterogeneity by destination income group (Table 9):
  - Positive relationship holds in AEs and EMDEs, stronger in EMDEs.
  - Example: For Middle-income Countries — Number of climate policies (in logs), destination country — Flows (column (5)): 0.2794* (standard error: 0.1641)
  - Interpretation: in EMDEs, investors may transfer firm-specific know-how and personnel, reducing absorptive capacity constraints.
- Source-country climate policies and outward FDI (Table 10):
  - Number of climate policies (in logs), source country — Green Flows (column (1)): 0.4188 (standard error: 0.3014)
  - Number of climate policies (in logs), source country — Green Projects (column (2)): 0.4878* (standard error: 0.2812)
  - Source-country policies also associated with higher bilateral non-green FDI flows/projects, suggesting possible complementary investments abroad or off-shoring of dirty activities.

### Policy instrument heterogeneity (section 3.2.1 and Table 11)
- Approach: re-balancing shares of policy types (revenue, expenditure, regulations) away from non-regulatory neutral policies while holding total policy count constant; coefficients π_p quantify the effect.
- Destination-country findings (Table 11, selected coefficients):
  - Share of government expenditure measures, destination country — Flows (column (2)): 3.0298*** (standard error: 1.0312)
  - Share of regulation measures, destination country — Flows (column (2)): 4.2384* (standard error: 2.3614)
  - Share of government revenue measures, destination country — Flows (column (2)): 6.2569*** (standard error: 1.9935)
  - Interpretation: re-balancing toward binding, budget-affecting measures (revenue and expenditure) and regulations is associated with larger green FDI inflows and projects.
- Source-country shares:
  - Number of climate policies (in logs), source country — Flows (column (5)): 0.7913*** (standard error: 0.3064)
  - Share of government expenditure measures, source country — Flows (column (5)): -4.3027** (standard error: 1.7972)
  - Interpretation: for source countries, re-balancing away from non-binding measures (especially toward expenditure measures) associated with lower bilateral flows, indicating potential trade-offs.

### EPS index robustness (Table 12)
- EPS measures (source country) and sub-components:
  - EPS index, Source country — Flows (column (1)): 0.2448** (standard error: 0.1197)
  - EPS Market based sub-component — Flows (column (5)): 0.3710*** (standard error: 0.1144)
  - Findings: higher EPS (more stringent policies) in source country associated with larger green FDI outflows/projects.
  - Breakdown: higher R&D subsidies in source country linked with lower green FDI outflows; taxes and certificates associated with higher green FDI outflows. Revenue measures as defined in Annex B do not show a statistically significant effect in CPD-based shares—differences may reflect EPS capturing intensity and a narrower set of revenue measures.
  - Caveat: analyses mostly capture short-term effects of subsidies; subsidies important for development of new LCTs and future deployment; trade-off between short- and long-term objectives noted. Feed-in tariffs show positive, marginally significant effect on green FDI outflows.

### Sectoral drivers in bilateral evidence
- Positive link between climate policies and green FDI driven by projects in industry, energy, and services.
- Industry: climate policies associated with increases in both green and non-green FDI (complementarity).
- Energy: climate policies associated with higher green and lower non-green projects (substitution).

### Overall conclusions and policy implications (Section 4)
- A higher number of climate policies—especially revenue and expenditure (budget-affecting) measures and regulations—are linked with higher green FDI inflows.
- Estimated effects on non-green FDI and on overall FDI are small and often statistically insignificant; historical aggregate effect on total greenfield investment is relatively negligible.
- IV evidence suggests OLS may understate the true effect (IV coefficients larger than OLS).
- Link between climate policies and green FDI inflows is highest in EMDEs, where technology diffusion via FDI is particularly relevant.
- Policy tensions and trade-offs:
  - Subsidies in source countries linked with lower green FDI outflows—illustrates tension between domestic support and global diffusion objectives.
  - Source-country climate policies can be associated with higher non-green outward FDI (possible off-shoring or complementary investments); further research needed to identify dominant mechanism.
- Policy recommendations highlighted by the evidence:
  - EMDEs have room to reduce tariffs on LCT goods to lower input costs and support green FDI.
  - Moving toward binding, budget-affecting climate policies (revenue and expenditure) and regulations tends to be more effective in attracting green FDI to destination countries.
  - International coordination in policy design can facilitate LCT deployment via FDI and reduce risk of offshoring polluting activities from AEs to EMDEs.

*Source: wpiea2024046-print-pdf - Climate Policies as a Catalyst for Green FDI — Working Paper No. WP/2024/046*

### section 4 concludes.

### section 4 concludes.

### Data
- Greenfield FDI data:
  - Source: fDi Markets database.
  - Coverage starts in 2003 and is updated monthly.
  - Data covers new projects and expansions of existing projects; collected primarily from public sources (including newswires from tens of thousands of global media sources and over 3000 promotion agency sources) and from market research and publication companies.
  - Projects cross-referenced against multiple sources, especially investing firms’ sources.
  - Excludes mergers and acquisitions and other equity and non-equity investments; joint ventures included if they lead to a new physical operation and if the project is majority owned by a foreign firm.
  - Combines announcements and opened projects, and includes multi-year investment plans (actual flows in a given year can be overreported).
  - In some instances investment figures are not provided and the database reports an estimated investment amount.
  - Validation notes: Strong correlation between country-level gross FDI flows and aggregate greenfield FDI values from fDi Markets (Aiyar, Malacrino and Presbitero (2023)); number of bilateral (country-pair) projects and investment values are highly correlated.
  - Project-level detail allows distinguishing types of investments; data classifies projects according to clusters and also tags projects.
  - The author creates a “green” label using clusters and tags (see Annex A); all other projects are classified as non-green (non-green can include direct substitutes of green projects or projects complementary or unrelated to green activities).
  - Sample restriction: countries that have at least one project (of any type) in more than 15 years.
  - Pattern: green FDI accelerated since 2016; green FDI flows accounted for 10 percent of total greenfield FDI between 2014 and 2017, and by 2022 it had reached 40 percent of total investment.  A similar increase is seen for green FDI projects as a percent of total projects.  A large amount of green FDI inflows into emerging market and developing economies still comes from advanced economies, although inflows from other emerging market and developing economies are not negligible.

- Aggregate FDI data:
  - Source: Financial Flows and Analytics (FFA) database constructed by the IMF’s Research Department.
  - Coverage: information for 165 countries dating back to 1970.
  - Compiles data on capital flows from the IMF’s Balance of Payments Statistics database and extends it with data from other sources including Haver Analytics, the CEIC and EMED databases.

- Climate policies:
  - Main source: Climate Policy Database (CPD).
  - CPD provides the most comprehensive international dataset on climate policies, although it is not exhaustive.
  - Database based on other international datasets, reports and country-specific documents; incorporates other databases such as the Climate Change Laws of the World and the OECD policy instruments database.
  - Considered generally complete for G20 economies (including EU member countries that are individual members of the G20, but not other EU members) and 18 other countries; includes advanced and emerging economies in Europe, Asia and Latin America and some less-developed countries.
  - Policy selection: only include policies that have climate change mitigation as one of their objectives (roughly 93 percent of policies).
  - EU policies applied to each member country’s policy portfolio; if a country became a member after the policy was decided in the EU, the date of policy adoption is the year of joining the EU.
  - Exclude sub-national policies.
  - Main variable: change in a country’s total number of active climate policies (and in some exercises subsets of policies).
  - CPD limitation: does not contain information about stringency of a country’s climate policy portfolio; OECD’s environmental policy stringency index (EPS) used as robustness (EPS has more limited country, sectoral, and instrument coverage).
  - Every policy in CPD carries information on policy objectives, administrative level, and instrument types.
  - Policies classified by budget impact into four categories:
    - (i) policies that generate revenue (such as carbon taxes or schemes capping emissions),
    - (ii) policies that generate expenses (e.g., R&D subsidies or feed-in-tariffs),
    - (iii) regulations with no budget impact,
    - (iv) nonregulatory budget neutral policies (e.g., national strategies, voluntary emission restrictions).
  - Timing and diffusion: climate policies accelerated in high-income countries following the Kyoto Protocol and the third IPCC assessment report; around the fourth assessment report the process sped up in middle-income and low-income countries.
  - Differences by income group: budget-neutral measures are most common in all countries; almost one-fifth of policies in advanced economies generate government expenditure (compared with just over 15 percent and 10 percent in middle-income and low-income countries, respectively).  Revenue-generating measures are used more frequently in advanced economies and, to a lesser extent, in middle-income countries.

- Country-level macroeconomic variables:
  - Capital stocks, employment, population and real GDP come from the Penn World Tables, version 10.1.
  - Total trade over GDP, average applied tariffs and average MFN tariffs come from the World Bank’s World Development Indicators (WDI).
  - Tariffs applied on low carbon technologies (LCT) goods (as defined in Howell et al. (2023)) constructed using tariff information from UNCTAD’s TRAINS; more details in Pienknagura (2024).

- Bilateral variables:
  - Data on distance, a trade agreement dummy, and other gravity variables from CEPII’s gravity database version 202211.

### Econometric approach
- Levels of aggregation exploited: aggregate country-year, sectoral, and bilateral (country-pair) analyses.

- Aggregate FDI inflows (country-year):
  - Aggregate greenfield FDI data at the recipient country-year level.
  - Distinguish between total greenfield FDI inflows (projects), green FDI inflows (projects), and non-green inflows (projects).
  - Two sets of regressions:
    - Relationship between climate policies and the level of FDI inflows (number of projects).
      - Use poisson-pseudo maximum likelihood estimator (PPML) to address zero values (Santos-Silva and Tenreyro (2006)).
      - Baseline equation:
        - y_h,i,t = exp{α_i + β log(CP_i,t−1) + γ X_i,t−1} + ε_i,t
        - where h ∈ {total, green, non−green}, y_h,i,t is either real dollar value of greenfield FDI inflows or number of projects of type h in country i in year t, α_i is a country fixed effect, log(CP_i,t−1) is natural logarithm of the stock of climate policies, X_i,t−1 includes controls in t−1: trade over GDP, log of the capital stock per employee, log of GDP per capita, and GDP growth.
      - Interest focuses on coefficient β. Extensions include country-specific time trends and lagged values of the left-hand-side variable.
    - Effect of climate policies on FDI inflows as a share of GDP (useful for comparison with previous studies).
      - Equation:
        - FDI_h,i,t / GDP_i,t = α_i + β log(CP_i,t−1) + γ X_i,t−1 + ε_i,t
      - Estimated using standard panel regression methods; h includes net total FDI flows from FFA when applicable.

- Sectoral FDI inflows:
  - Use sector target information from fDi Markets to construct sector-level FDI inflows and number of projects.
  - Study heterogeneous effects across sectors via:
    - y_h,i,s,t = exp{α_i + ω_i * year + σ_s,t + γ X_i,t + Σ_s φ_s log(CP_i,t) 1(sector = s)} + ε_i,s,t
    - φ_s is sector-specific elasticity of FDI inflows (projects) of flow type h with respect to climate policies; 1(sector = s) indicator equals 1 if project targets sector s.
    - Specification includes sector-time fixed effects σ_s,t, country fixed effects α_i, and country-specific time trend ω_i * year.
  - Sector mapping: fDi Markets sectors mapped to ISIC one digit sectors and grouped into Agriculture and Mining, Energy, Construction, Manufacturing, Services, and Others (Others includes Waste management and Space and Defense).

- Bilateral FDI flows (gravity-style):
  - Construct bilateral FDI flows x_h,i,j,t from sender country j to recipient country i in year t for FDI type h.
  - Use PPML estimator to handle zeroes:
    - x_h,i,j,t = exp{α_i + ω_j,t + β log(CP_i,t−1) + γ X_i,t−1 + δ Z_i,j,t−1} + ε_i,t
    - Include country i average applied tariff on LCT goods and average applied tariff on merchandise imports, country-pair variables (geographic distance, trade agreement dummy). Some specifications use country-pair fixed effects instead of time-invariant bilateral variables.
    - x is either bilateral FDI dollar flow (real terms) or number of projects of type h. Sender-country time-varying variables are captured by ω_j,t.
  - Also study policy spillovers by swapping roles (i as sender, j as destination) and controlling for destination country-year fixed effects.
  - Policy instrument heterogeneity:
    - Group policies by budget impact: revenue, expenditure, regulations, and non-regulatory neutral (excluded in some specifications).
    - Extended specification:
      - x_h,i,j,t = exp{α_i + ω_j,t + β log(CP_i,t−1) + γ X_i,t−1 + δ Z_i,j,t−1 + Σ_p π_p share_i,p,t−1} + ε_i,t
      - share_i,p,t−1 is the share of climate policies of type p ∈ {revenue, expenditure, regulations} in country i’s climate policy portfolio.
      - Because the (log) of country i’s climate policy portfolio is controlled and non-regulatory neutral policies are excluded, π_p captures effect of increasing share of type p at the expense of non-regulatory neutral policies while keeping the size of the climate portfolio constant.
      - Specification helps when many countries have no policies of some types (log transformation would drop them).
    - Elasticity formula for green FDI with respect to policy type p:
      - ∂ ln FDI_green / ∂ ln P_p = β share_p + π_p {(1 − share_p) share_p − share_p Σ_{q ≠ p} π_q share_q}
      - Using (6), effect of each policy type on green FDI flows estimated for country with average share of each policy type.

### Results (overview of approach to be presented in section 3)
- Analysis sequence:
  - Begin with evidence based on aggregate FDI flows.
  - Exploit sectoral information to explore heterogeneous effects across sectors.
  - Present evidence from bilateral FDI flows, focusing on connections with climate policies implemented in both source and destination countries.

*Source: wpiea2024046-print-pdf - section 4 concludes.*

### 3.1    Evidence from Aggregate FDI inflows

### 3.1    Evidence from Aggregate FDI inflows

### Main empirical findings
- A higher number of active climate policies is associated with larger green FDI inflows and projects.  
- Estimated coefficients for a country’s climate policy count are positive and statistically significant in specifications for both inflows and projects, with the coefficient for projects similar to that for inflows (Tables 1; columns (1) and (5)).
- Inclusion of a country-specific linear trend leaves the magnitude and statistical significance of the climate-policy coefficient virtually unchanged (Tables 1; columns (2) and (6)).
- Controlling for persistence in FDI by adding the (log of) past FDI flows/values reduces the point estimate for climate policies but the coefficient remains statistically significant (Tables 1; columns (3) and (7)).
- Re-estimating specifications on the sub-sample excluding observations where flows/projects int−1 are zero yields virtually unchanged results compared to the full sample (Tables 1; columns (4) and (8)), suggesting sample selection is not driving differences between specifications.

### Economic magnitude
- A one standard deviation increase in the number of climate policies is associated with a 15 percent increase in green FDI inflows and projects.
- Using the most conservative estimates (Tables 1; columns (3) and (7)), an increase in the climate policy count from the EMDE median in 2019 to the 75th percentile would be related with a 20 percent increase in green FDI flows/projects.

### Effects on non-green and total FDI
- Estimating the same equation for non-green FDI flows/projects shows either a positive and significant association or a statistically insignificant relationship, depending on specification (Table 2).
- Estimated coefficients for non-green FDI are substantially lower than those for green FDI inflows.
- As a result, the link between climate policies and total FDI inflows is in most cases positive and statistically significant, with the estimated coefficient smaller than for green FDI flows (Table 3).
- When expressing FDI flows as a share of GDP (equation (2); Table 4), a higher number of climate policies is robustly associated with higher levels of green FDI as a share of GDP (columns 1–4). By contrast, columns 5–8 suggest climate policies are associated with lower non-green FDI inflows as a share of GDP, although the coefficient is statistically insignificant.
- Results for overall greenfield FDI as a share of GDP point to a statistically insignificant relationship between climate policies and overall greenfield FDI as a share of GDP (Table 5).
- Similar conclusions emerge for net aggregate FDI inflows (column 5), consistent with prior findings.

### Robustness and endogeneity (instrumental variables)
- Concerns that countries with larger green FDI stocks may also have more active climate policies are addressed via an IV exercise where climate policies are instrumented by the distance-weighted stock of policies in other countries (Table 6).
- Reassurance for the exclusion restriction: the coefficient for distance-weighted policies abroad, in a regression of green FDI that also includes domestic policies, is not statistically significant (Table 6; columns 1 and 2).
- IV results confirm the positive link between climate policies and green investments as a percent of GDP, with the estimated IV coefficient being approximately twice as large as the OLS results in Table 4 (Table 6; columns 3 and 4), suggesting anticipation effects and endogeneity tend to dampen the observed OLS relationship.
- The Kleibergen-Paap Wald statistic is above the Cragg-Donald critical values, suggesting that the null of weak instruments is rejected.

### Sectoral heterogeneity
- Allowing the climate-policy coefficient to vary across sectors (equation 3; Table 7) reveals heterogeneity:
  - Positive and statistically significant relationships between climate policies and green FDI projects and flows into industry and service sectors.
  - Positive and statistically significant association with green FDI inflows into the energy sector (for projects the coefficient is positive but statistically insignificant).
  - For other sectors, estimated coefficients are insignificant and in some cases negative.
- Adverse relationships between climate policies and non-green FDI flows are concentrated in a handful of sectors:
  - Largest adverse effects on non-green FDI flows and projects into agriculture and energy.
  - Associations with lower non-green construction projects.
  - Industry is notable for showing a positive and significant relationship between climate policies and non-green FDI inflows.
- Interpretation:
  - Industry shows complementarities between green and non-green activities (e.g., inputs used by both green and non-green production).
  - Energy shows a clearer substitution pattern: stronger climate agendas are associated with higher green projects at the expense of non-green ones.

### Overall conclusion
- Tables 1–6 document a systematic positive and significant relationship between countries’ number of active climate policies and green greenfield FDI inflows.
- The link with non-green and overall greenfield FDI is either small (in levels) or statistically insignificant (as a share of GDP), resulting in a relatively negligible aggregate effect on total greenfield investment historically.
- Results are robust to trends, lagged dependent variables, sample checks, and an IV strategy; IV estimates suggest OLS may understate the true effect.
- These patterns are based on historical data and may change as the balance between green and non-green FDI evolves.

*Source: wpiea2024046-print-pdf - 3.1    Evidence from Aggregate FDI inflows*

### 3.2    Evidence from Bilateral Flows

### 3.2    Evidence from Bilateral Flows

### Methodology and identification
- Uses bilateral gravity estimations based on the specification in equation 4 to study the link between climate policies and bilateral green and non-green FDI flows.
- Key controls and fixed effects:
  - Destination country fixed effects (capture time-invariant country characteristics).
  - Source country-year fixed effects (capture source country-specific variables such as growth, rule of law, or level of development).
  - Dummy for country-pair trade agreement.
  - Either standard time-invariant gravity variables (distance, common language) or a full set of country-pair fixed effects.
  - Additional controls: log of (real) GDP and population (size), log of destination country capital stock (diminishing returns to capital), and trade-weighted LCT and overall applied tariffs.
- Hybrid gravity approach is also applied when studying climate policies in the source country.
- Limitations: cannot include full set of destination-time fixed effects because destination climate policies do not vary across source countries.

### Evidence on destination-country climate policies and green FDI
- A larger number of climate policies in the destination country is associated with higher bilateral green FDI flows and higher number of green projects (reported in Table 8, Columns 1 and 3).
- Magnitude: a one standard deviation increase in the stock of climate policies in the recipient country is associated with an increase in bilateral green FDI inflows (the number of green projects) of 7 percent (6 percent), on average.
- Estimated coefficients are robust to inclusion of a full set of country-pair dummies (Table 8, columns 2 and 4).
- No systematic evidence that destination-country climate policies hinder non-green FDI inflows (Table 8, columns 5 and 6); there is a negative link with bilateral non-green projects (columns 7 and 8) but coefficients are smaller than those for green projects.

### Trade openness and LCT deployment
- Lower tariffs on LCT goods (trade openness) are associated with higher green FDI inflows and more green projects.
- Interpretation: trade protection can induce tariff-jumping FDI but also raise input costs; results point to the latter effect dominating for green FDI—lower LCT tariffs reduce technology costs and support LCT diffusion via FDI.
- Policy implication: EMDEs in particular have substantial room to reduce tariffs on LCT goods.

### Heterogeneity by destination income group (AEs vs EMDEs)
- Sample split between advanced economies (AEs) and emerging market and developing economies (EMDEs) (see Table 9).
- Positive relationship between destination climate policies and green FDI inflows/projects holds in both income groups, and is stronger in EMDEs:
  - In EMDEs, coefficients are statistically significant in all cases.
  - In AEs, effects are significant only for green projects and estimated coefficients are somewhat smaller.
- Interpretation: foreign investors may be less constrained by absorptive capacity limitations in EMDEs because they can transfer firm-specific know-how and deploy qualified personnel to affiliates.

### Source-country climate policies and outward FDI
- Modified specification of equation 4: controls for source country fixed effects, source country-time varying variables (log of GDP and population), and a full set of destination country-year fixed effects.
- Results (Table 10):
  - Climate policies in the source country are linked with higher bilateral green FDI flows and projects; coefficients are larger and more statistically significant for green projects (Table 10, columns 3, 4).
  - For green flows, the effect is statistically significant only when controlling for the full set of country-pair fixed effects (Table 10, column 2).
  - Climate policies in the source country are also associated with higher bilateral non-green FDI flows, especially for projects (Table 10, columns 5–8).
- Possible mechanisms for non-green outward FDI increase:
  - Complementary green and non-green investments abroad used to comply with source-country regulations (e.g., vehicle emissions regulations).
  - Off-shoring of “dirty” projects in response to stricter domestic regulation.
- Recommendation: further research needed to identify which mechanism dominates.

### International coordination and cooperation
- Bilateral flow results highlight the importance of international coordination:
  - Concerted climate policies may lower risk of offshoring polluting activities from AEs to EMDEs.
  - Coordination in policy design can facilitate LCT deployment to EMDEs via green FDI.

### Heterogeneity by policy instrument (section 3.2.1)
- Uses policy typology from Annex B: distinguishes revenue measures, expenditure measures, regulations, and non-binding policies; classification based on budget impact and whether policies are binding.
- Estimation approach based on equation 5:
  - Coefficients π_p quantify change in green FDI flows/projects as countries re-balance policy portfolios from non-binding to binding measures, holding total number of policies constant.
  - Allows computation of marginal effect of an additional policy type using equation 6.
- Key findings (Table 11; Figure 3):
  - Destination-country re-balancing from non-binding policies towards revenue-generating measures and expenditure measures is associated with increases in both total green FDI inflows and projects.
  - Re-balancing towards regulation also boosts green FDI, though effect on flows is weaker statistically.
  - Conclusion: moving toward binding policies appears more effective in attracting green FDI.
  - For source countries, re-balancing away from non-binding measures is associated with lower bilateral flows, particularly for expenditure measures.

### EPS index and source-country policy stringency (Table 12)
- Uses OECD EPS index and sub-components: (i) taxes and certificates, (ii) non-market based policies (regulations), (iii) feed-in-tariffs, (iv) R&D subsidies. Notes on mapping:
  - EPS components roughly map to revenue measures, regulations, and expenditure measures.
  - EPS has narrower policy set and smaller country coverage than CPD but includes many source countries.
- Findings:
  - Higher EPS values in the source country (more stringent climate policies) are associated with larger green FDI outflows/projects abroad (Table 12, columns 1, 2, 5, 6).
  - Breakdown: higher R&D subsidies in the source country are linked with lower green FDI outflows.
  - Taxes and certificates (EPS measure) are associated with higher green FDI outflows, whereas revenue measures as defined in Annex B do not show a statistically significant effect.
  - Possible reasons for divergence: EPS captures policy intensity and focuses on a narrower set of revenue measures.
- Caveat: analyses capture mostly short-term effects of subsidies; subsidies may be important for development of new LCTs and future deployment. Trade-off between short- and long-term deployment objectives is noted. Feed-in tariffs show a positive, albeit marginally significant, effect on green FDI outflows.

### Sectoral heterogeneity
- Positive link between climate policies and green FDI is driven by projects in industry, energy, and services.
- Sectoral differences:
  - In industry, climate policies are associated with increases in both green and non-green FDI (complementarities).
  - In energy, climate policies hamper non-green projects.
- Conclusion: effects vary across sectors and within-sector activities.

### Overall conclusions and policy implications (Section 4)
- A higher number of climate policies—especially revenue and expenditure (budget-affecting) measures and regulations—are linked with higher green FDI inflows.
- Estimated effects on non-green FDI and on overall FDI are small and statistically insignificant.
- Implication: economic costs of the climate transition, from the perspective of FDI, appear to be small.
- Link between climate policies and green FDI inflows is highest in EMDEs, where diffusion of technologies is particularly relevant.
- Some policy tensions:
  - Subsidies in the source country are linked with lower green FDI outflows, illustrating potential tension between domestic support and global diffusion objectives.
  - Importance of international coordination and cooperation in policy design to facilitate LCT deployment via FDI and to avoid undesirable spillovers such as offshoring of polluting activities.

*Source: wpiea2024046-print-pdf - 3.2    Evidence from Bilateral Flows.*

### References

### References (Content Unit)

### Overview of source material
- Contains the list of cited literature used in the chapter, followed by three appendices:
  - Appendix A: Identifying “Green” Greenfield FDI Projects (classification and counts).
  - Appendix B: Classification of Climate Policies by their Impact on the Government’s Budget (mapping of CPD categories to budget impact).
  - Appendix C: Figures and Tables (empirical results and robustness checks across multiple specifications).

### Appendix A — Identifying “Green” Greenfield FDI Projects
- Classification rule:
  - All projects in the “Environmental Technology” Cluster are labeled as green.
  - Projects with any of the following tags are also labeled as green:
    - Alternative proteins
    - Carbon capture
    - Cleantech
    - Cultured meats
    - Electric vehicles
    - Hydorgen
    - Photovalic
    - Plant-based foods
    - Vegan industries
    - Wind power technologies
    - Sustainable tourism
    - Waste to energy
- Coverage and counts:
  - Labeled about 13,000 projects as green over the 20 years of available data, out of a total of close to 300,000 (about 4 percent).

### Appendix B — Classification of Climate Policies by Impact on Government’s Budget
- Policy mapping categories:
  - Expense (generates government expenses): Direct investment; Funds to sub-national governments; Infrastructure investment; Demonstration projects; Research program; Technology development; Technology deployment and diffusion; Feed-in-tariffs or premiums; Loans; Grants and subsidies; Retirement premium; Tax relief.
  - Revenue (generates government revenue): Removal of fossil fuels; CO2 taxes; Energy and other taxes; User charges; GHG emission reduction crediting and offsetting; GHG emission allowance.
  - Neutral, regulations (budget-neutral but regulatory compliance costs): Grid access and priority for renewables; Performance label; Institutional creation; Strategic Planning; Auditing; Codes and standards; Building Standards; Industrial air pollution standards; Product standards; Sectoral standards; Vehicle air pollution standards; Vehicle fuel-economy and emission standards; Monitoring; Obligation schemes; Other mandatory requirements.
  - Neutral, non-regulatory (budget-neutral and non-regulatory): Formal and legally binding climate strategy; Political and non-binding climate strategy; Procurement rules; Tendering schemes; Green and white certificates; Advice or aid in implementation; Information provision; Performance label; Comparison label; Endorsement label; Professional training and qualification; Institutional creation; Strategic planning.

### Appendix C — Figures and Tables: Key empirical findings and statistics
- Figures:
  - Figure 1: Evolution and composition of green FDI flows and projects (percent of total) for 2003–2021 and 2015-2022 averages by income group. Source: Hasna et al. (2023) based on the FT's fDi markets database.
  - Figure 2: Evolution and composition of climate policies across income groups (1990–2020 series shown; 2015-2021 averages by income group).
  - Figure 3: Impact of climate policies, by policy type (response to a one st. dev. change in each instrument; whiskers are 90 percent confidence interval).
- Selected table findings (preserve reported coefficients and significance exactly as in the source):
  - Table 1: Climate Policies and Green FDI: Evidence from Aggregate Data
    - log Number of climate policies (t-1) — Projects (column (1)): 0.6358*** (standard error: 0.1146)
    - log Number of climate policies (t-1) — Inflows (column (5)): 0.6341*** (standard error: 0.1400)
  - Table 2: Climate Policies and Non-Green FDI: Evidence from Aggregate Data
    - log Number of climate policies (t-1) — Projects (column (1)): 0.1682*** (standard error: 0.0351)
    - log Number of climate policies (t-1) — Inflows (column (6)): 0.0971* (standard error: 0.0525)
  - Table 3: Climate Policies and Total Greenfield FDI: Evidence from Aggregate Data
    - log Number of climate policies (t-1) — Projects (column (1)): 0.1790*** (standard error: 0.0361)
    - log Number of climate policies (t-1) — Inflows (column (6)): 0.1370*** (standard error: 0.0526)
  - Table 4: Climate Policies and Greenfield FDI (as a share of GDP)
    - log Number of climate policies (t-1) — Green FDI (column (1)): 0.1159*** (standard error: 0.0406)
    - log Number of climate policies (t-1) — Non-Green FDI (column (5)): -0.3366 (standard error: 0.2153) (noted as not significant at conventional levels)
  - Table 5: Climate Policies and FDI (as a share of GDP)
    - GDP growth (t-1) coefficients: 7.5729*** (column (1), standard error: 2.5917) and other growth coefficients reported across columns.
  - Table 6: Climate Policies and Greenfield FDI (as a share of GDP)—IV Results
    - OLS (column (1)): log Number of climate policies (t-1) = 0.1628** (standard error: 0.0821)
    - IV (column (2)): log Number of climate policies (t-1) = 0.2222** (standard error: 0.1058)
    - IV (column (4)): log Number of climate policies (t-1) = 0.4376** (standard error: 0.2224)
  - Table 7: Sector Level Data (selected interactions)
    - log Number of climate policies (t-1) — Green Projects (column (1)): 0.2183** (standard error: 0.0910)
    - Interaction: log Number of climate policies (t-1) x industry dummy — Green Projects (column (5)): 0.2616*** (standard error: 0.0929)
    - Interaction: log Number of climate policies (t-1) x energy dummy — Non-green Projects (column (3)): -0.2070*** (standard error: 0.0339)
  - Table 8: Climate Policies in the Destination Country and Greenfield FDI
    - Number of climate policies (in logs), destination country — Green Flows (column (1)): 0.2852* (standard error: 0.1548)
    - Trade-weighted LCT tariffs, destination country — negative and significant coefficients (e.g., column (1): -0.1091***, standard error: 0.0160)
  - Table 9: Destination country results by income level (selected)
    - For Middle-income Countries — Number of climate policies (in logs), destination country — Flows (column (5)): 0.2794* (standard error: 0.1641)
  - Table 10: Climate Policies in the Source Country and Greenfield FDI
    - Number of climate policies (in logs), source country — Green Flows (column (1)): 0.4188 (standard error: 0.3014)
    - Number of climate policies (in logs), source country — Green Projects (column (2)): 0.4878* (standard error: 0.2812)
  - Table 11: Role of Different Policies (destination and source country shares)
    - Share of government expenditure measures, destination country — Flows (column (2)): 3.0298*** (standard error: 1.0312)
    - Share of regulation measures, destination country — Flows (column (2)): 4.2384* (standard error: 2.3614)
    - Share of government revenue measures, destination country — Flows (column (2)): 6.2569*** (standard error: 1.9935)
    - Number of climate policies (in logs), source country — Flows (column (5)): 0.7913*** (standard error: 0.3064)
    - Share of government expenditure measures, source country — Flows (column (5)): -4.3027** (standard error: 1.7972)
  - Table 12: Robustness to EPS index (source country)
    - EPS index, Source country — Flows (column (1)): 0.2448** (standard error: 0.1197)
    - EPS Market based sub-component — Flows (column (5)): 0.3710*** (standard error: 0.1144)
- Notes on statistical reporting across tables:
  - "Robust Standard Errors in parenthesis."
  - Significance notation: *** p < 0.01, ** p < 0.05, * p < 0.1
  - Fixed effects and trends are reported per table (e.g., Country FE YES; Country-specific trend YES/NO; Destination country FE YES; Source country-Year FE YES; Source country-Destination country FE YES/NO).
  - Observations and R-squared values are reported for each specification (examples across tables include Observations: 1608, 1608, 999, 999; 1,660; 2,088; 39,719; 111,359; 57,722; 160,501; 2,041; 1,964; etc.; R-squared examples: 0.3415, 0.3606, 0.2477, 0.3101, 0.3163, 0.4133).

*Climate Policies as a Catalyst for Green FDI — Working Paper No. WP/2024/046*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024046-print-pdf.pdf_