No Plan, No Aid?

Exploring the Impact of National Adaptation Plans on Adaptation Finance

By Pierre Beaucoral, Michaël Goujon, and Sébastien Marchand

Download Full Paper Supplementary Material Source Code

Summary

“No Plan, No Aid?” examines whether implementing a National Adaptation Plan (NAP) increases a country’s climate adaptation finance. Through a novel theoretical model and a robust Double Machine Learning (DML) approach, the authors find that NAP adoption does send mixed signals—reducing perceived vulnerability while boosting perceived capacity. Crucially, they find that while traditional econometric models show no impact or even a negative one, modern methods reveal a significant positive effect on adaptation aid.

Key Insights

📊 Donor Strategy

Donors balance between rewarding vulnerability (Donor V) and rewarding institutional capacity (Donor C), shaping who gets aid post-NAP.

🤖 DML Advantage

Double Machine Learning reveals significant causal effects that traditional methods overlook due to confounding and endogeneity.

📈 Measurable Impact

DML ATE estimate: +56.75 million USD; IV-DML LATE: +311 million USD. Strong effects for “compliers” near NAP threshold.

🧭 Strategic Recommendation

NAP adoption is especially effective in low-revenue, low-trade, or low-GDP contexts. Prioritize capacity-building to unlock funding.

Methods & Results Summary

The paper builds a game-theoretic model contrasting two donor types: Donor V (vulnerability-focused) and Donor C (capacity-focused). Countries signal institutional capacity by adopting a NAP, but may lose aid from Donor V as they appear “less needy.”

Empirically, authors apply OLS, 2SLS (using latitude as an instrument), and Double Machine Learning (DML) to a dataset of 2,100 observations across 2013–2023.

  • 📉 OLS: No significant NAP effect.
  • 🧪 2SLS: Weakly positive effect (large std. error).
  • 🤖 DML (ATE): +56.75 (p = 0.04)
  • 🚀 IV-DML (LATE): +311.1 (p = 0.03)

Results reveal a hidden effectiveness of NAPs—detectable only when robust methods correct for nonlinearity and omitted variables.