// Academic Project · Year 4
Data Science: What Drives Public Support for Climate Action?
Modelled support for climate action across 62 countries and found trust in institutions, not worry, divides them.
Group coursework for Data Science, using the UN Development Programme's Peoples' Climate Vote 2024, a survey of public opinion on climate change around the world. We looked at whether age changes how urgently people want to move to renewable energy, modelled what predicts support for stronger climate commitments, and grouped countries by their attitudes.
Overview
- Module
- COMP5122M Data Science
- Year
- 2025/26 · Year 4
- Team
- 4 students
- Weighting
- Two reports, 18% and 32%
Both parts used Python with pandas, scikit-learn, matplotlib and seaborn, and each ended in a short written report.
Is there an age gap?
Part 1: exploratory analysis
The survey asked how quickly each country should replace coal, oil and gas with renewables. We turned the answers into a 1 to 5 urgency score, weighted by the survey's own weights, and compared age groups country by country across 45,784 records.
- • People aged 18 to 35 and 36 to 59 ranked countries almost identically (a correlation of 0.92 across 62 countries).
- • People over 60 were a little less urgent, by about 0.2 points, but the survey only had data for them in 15 to 17 countries, so that gap is the least certain finding.
- • The biggest age gaps were in the USA (0.45) and the UK (0.27). In the USA, people aged 36 to 59 scored far lower than the global pattern predicts.
What predicts support
Part 2: supervised learning
We built a dataset of 124 country and age-group combinations across 62 countries, with 21 features from the survey, and predicted a weighted score for support for stronger climate commitments. We chose regression over the suggested classification, to keep the detail in the score.
- • A random forest explained 64% of the variance on the test set (R² 0.642), slightly ahead of linear regression (0.624). With so few rows, cross-validated scores were lower, 0.50 for the random forest.
- • Wanting rich countries to help poorer ones was the strongest predictor in the linear model. In the random forest, wanting protection from extreme weather mattered most, at 19% of feature importance.
Grouping countries
Part 2: clustering
We clustered the 62 countries with k-means into four groups and checked the grouping with hierarchical clustering. Two principal components captured 58% of the variation, enough to plot the groups. Worry about climate change was high in every group. What split them was trust in the institutions meant to act on it:
- • Emerging giants (China, India, Indonesia) had the highest trust in government and business.
- • The sceptical West (the USA, UK, Germany, Australia and others) had the lowest trust in government and the weakest support for stronger commitments.
- • The engaged Global South, mostly Latin America, was the most worried, but shared the West's low trust.
- • Developing climate advocates, mainly in Sub-Saharan Africa and South Asia, gave the strongest support for moving to renewables.
