OLS regression analysis testing whether CO₂ emissions per capita and renewable energy share predict GDP growth across 38 OECD countries (2010–2020), using World Bank Sovereign ESG data.
| Hypothesis | Test | Statistic | p-value | Decision |
|---|---|---|---|---|
| H₁: CO₂ ↔ GDP Growth | Pearson r | r = 0.0575 | 0.2405 | Fail to reject H₀ |
| H₂: Renewable % ↔ GDP Growth | Pearson r | r = −0.0485 | 0.3225 | Fail to reject H₀ |
| H₃: Overall model significance | F-test | F = 0.8775 | 0.4166 | Fail to reject H₀ |
| H₄: High vs Low Renewable GDP | Welch's t-test | t = −2.0018 | 0.0460 | Reject H₀ ✅ |
OLS model: R² = 0.004, Adj. R² = −0.001 — environmental indicators alone explain less than 1% of GDP growth variance in OECD countries over this period.
| Property | Value |
|---|---|
| Source | World Bank Sovereign ESG Data Portal |
| Raw size | 16,969 rows × 68 columns (all countries, 1960–2023) |
| After filtering | 418 observations (38 OECD countries, 2010–2020) |
| Target indicators | CO₂ per capita (EN.ATM.CO2E.PC), Renewable % (EG.FEC.RNEW.ZS), GDP growth (NY.GDP.MKTP.KD.ZG) |
Descriptive statistics (n = 418):
| Variable | Mean | Std | Min | Max | Skewness |
|---|---|---|---|---|---|
| CO₂ (t/capita) | 7.44 | 3.87 | 1.38 | 21.76 | 1.14 |
| Renewable (%) | 21.98 | 16.13 | 1.30 | 82.90 | 1.50 |
| GDP Growth (%) | 1.86 | 3.16 | −10.94 | 24.62 | −0.16 |
| Figure | What it shows |
|---|---|
 |
Histograms + KDE for all three variables |
 |
Boxplots showing spread and outliers |
 |
OECD mean trends over time (2010–2020) |
 |
Pearson correlation matrix heatmap |
 |
Scatter plots: IV₁ and IV₂ vs GDP growth |
 |
OLS coefficient plot with 95% CIs |
 |
Residuals vs fitted + Q-Q plot |
 |
H₄ group comparison boxplot |
 |
Top 15 OECD countries by avg GDP growth |
 |
COVID-19 shock — country trajectories 2010–2020 |
flowchart LR
A["World Bank ESG CSV<br/>16,969 × 68"] --> B["Filter<br/>38 OECD · 2010–2020"]
B --> C["Feature prep<br/>CO₂ · Renewable % · GDP growth<br/>n = 418"]
C --> D["EDA<br/>distributions · correlations"]
C --> E["OLS + Pearson r<br/>H₁ · H₂ · H₃"]
C --> F["Welch t-test<br/>H₄ groups"]
E --> G["Diagnostics<br/>AD · BP · DW · VIF"]
D --> H["Figures 01–10"]
E --> H
F --> H
G --> H
- Ingest & filter — the bulk World Bank CSV is narrowed to 38 OECD countries, 2010–2020.
- Feature prep — three indicators are aligned into a 418-row modelling frame.
- Testing — OLS with Pearson correlations (H₁–H₃) and a Welch t-test for the renewable-share split (H₄).
- Diagnostics — Anderson–Darling, Breusch–Pagan, Durbin–Watson, and VIF checks.
- Output — ten figures rendered to
figures/.
- OLS regression (statsmodels) with two independent variables
- Pearson correlation for H₁ and H₂
- F-test for overall model significance (H₃)
- Welch's t-test for group comparison (H₄)
- Regression diagnostics: Anderson-Darling normality test, Breusch-Pagan homoscedasticity test, Durbin-Watson autocorrelation check, VIF for multicollinearity
Diagnostic results:
- Normality: Anderson-Darling statistic = 14.63, p < 0.001 — residuals deviate from normality
- Homoscedasticity: Breusch-Pagan statistic = 6.70, p = 0.035 — variance not constant
- Autocorrelation: Durbin-Watson = 1.49 — possible autocorrelation
- Multicollinearity: VIF = 1.14 for both predictors — no multicollinearity concern
Highest average GDP growth (2010–2020): Ireland (6.37%), Türkiye (5.50%), Lithuania (3.52%), Israel (3.46%), Poland (3.27%)
Lowest average GDP growth: Greece (−2.64%), Italy (−0.60%), Spain (−0.003%), Portugal (0.01%), France (0.60%)
COVID-19 impact: OECD mean GDP growth dropped from 2.25% in 2019 to −3.88% in 2020.
git clone https://github.com/kandulanikhilvarma/esg-gdp-regression.git
cd esg-gdp-regression
pip install -r requirements.txt
jupyter notebook ESG_Regression_Analysis.ipynbThe notebook expects
ESGCSV.csvin the same directory. If the file is too large for your clone, download the bulk CSV from the World Bank ESG Data Portal and rename itESGCSV.csv.
Source data is the World Bank Sovereign ESG Data portal, © World Bank Group, distributed under the Creative Commons Attribution 4.0 (CC BY 4.0) license. The World Bank does not endorse this analysis or its conclusions; the filtering, modelling, and interpretation here are entirely my own.
Analysis code in this repository is released under the MIT License (see LICENSE); the source data remains under its original terms.
- Angelidis et al. (2024). World ESG performance and economic activity. Journal of International Financial Markets, Institutions and Money. https://doi.org/10.1016/j.intfin.2024.101986
- Dkhili, H. (2019). Environmental performance and economic growth in MENA countries. Journal of Health and Pollution, 9(24), 191208. https://doi.org/10.5696/2156-9614-9.24.191208
- Mirziyoyeva & Salahodjaev (2023). Renewable energy, GDP and CO₂ emissions in high-globalized countries. Frontiers in Energy Research, 11, 1123269. https://doi.org/10.3389/fenrg.2023.1123269
- OECD (2024). Member Countries. https://www.oecd.org/about/members-and-partners/
- World Bank Group (2024). Sovereign ESG Data Portal. https://datacatalog.worldbank.org/search/dataset/0037651
Released under the MIT License — see LICENSE.
Nikhilvarma Kandula — Data Science · NLP · Statistical Analysis
LinkedIn · Email · Portfolio