Assessing the efficiency of renewable energy policies: A DEA, machine learning, and panel data approach

  • 16 Views
  • 3 Downloads

Creative Commons License DMCA.com Protection Status
This work is licensed under a Creative Commons Attribution 4.0 International License

Type of the article: Research Article

Abstract
The accelerating energy transition and uneven effectiveness of renewable energy policies across OECD member, accession, and key partner countries covered by the OECD policy database highlight the need to understand which policy designs deliver efficient outcomes. This study aims to assess how different regulatory policy instruments and their combinations influence the efficiency with which countries convert policy stringency inputs into renewable electricity generation outcomes. The analysis uses panel data for 47 countries over 2000–2023, combining Data Envelopment Analysis, Random Forest modeling, and fixed-effects regressions with lagged variables. The results show that the average DEA efficiency score is 0.348 (median = 0.258), indicating substantial underperformance relative to the best-practice frontier. Random Forest results reveal that structural factors dominate, with energy intensity (IncNodePurity = 2.53 × 10–12) and GDP per capita (1.44 × 10–12) exhibiting the highest variable importance scores. Among policy instruments, feed-in tariffs, renewable energy auctions, and emission standards have the most robust positive associations. Interaction effects indicate complementarity, with the joint impact of feed-in tariffs and auctions being positive and significant. In contrast, most individual instruments show limited or insignificant effects, confirming that policy effectiveness depends on coordinated policy mixes and structural conditions rather than isolated measures.

Acknowledgment
This article was prepared based on the results of the project 101127491- EnergyS4UA-ERASMUS-JMO2023-HEI-TCH-RSCH. Views and opinions expressed are, however, those of the authors only and do not necessarily reflect those of the European Union or European Education and Culture Executive Agency. Neither the European Union nor the granting authority can be held responsible for them.

view full abstract hide full abstract
    • Figure 1. The graphical representation of variable importance derived from the Random Forest model
    • Figure 2. Random Forest variable importance plot
    • Figure 3. Partial dependence plot
    • Table 1. Random Forest variable importance for DEA efficiency
    • Table 2. Fixed-effects quadratic model for feed-in tariffs and DEA efficiency
    • Table 3. Fixed-effects model for DEA efficiency with lagged policy variables
    • Table 4. Fixed-effects model with policy interaction effects
    • Table 5. Fixed-effects quadratic models for auctions and planning policies
    • Table A1. Descriptive statistics of variables (47 OECD member, accession, and key partner countries, 2000–2023)
    • Conceptualization
      Maksym W. Sitnicki, Serhiy Lyeonov, Dmytro Kurinskyi, Oleksii Havrylenko
    • Software
      Maksym W. Sitnicki, Oleksii Havrylenko
    • Visualization
      Maksym W. Sitnicki, Oleksii Havrylenko
    • Writing – original draft
      Maksym W. Sitnicki, Serhiy Lyeonov, Dmytro Kurinskyi, Oleksii Havrylenko
    • Writing – review & editing
      Maksym W. Sitnicki, Serhiy Lyeonov, Dmytro Kurinskyi, Oleksii Havrylenko
    • Project administration
      Serhiy Lyeonov
    • Supervision
      Serhiy Lyeonov
    • Funding acquisition
      Dmytro Kurinskyi
    • Resources
      Dmytro Kurinskyi
    • Data curation
      Oleksii Havrylenko
    • Formal Analysis
      Oleksii Havrylenko
    • Investigation
      Oleksii Havrylenko
    • Methodology
      Oleksii Havrylenko
    • Validation
      Oleksii Havrylenko