Introduction

Green bonds—debt issued to finance environmental projects—have grown rapidly. Do green bonds provide better risk-adjusted returns than traditional bonds, or is the green premium a speculative bubble? Machine learning attribution analysis reveals factors driving green bond outperformance and assesses sustainability of premium.

Attribution Analysis

Compare green bond returns to matched traditional bonds (same issuer, maturity, credit quality). Measure outperformance (alpha). Use ML to attribute alpha to: (1) supply/demand imbalances (ESG funds buying, supply limited); (2) selection bias (green issuers better credit quality); (3) true fundamental premium (environmental impact value). Separate permanent vs temporary sources.

Results

Green bond premium is partially due to selection bias (green issuers tend to be higher quality) and partially supply/demand. Fundamental premium (value from avoided environmental damages) is harder to quantify but seems limited. Green bonds offer competitive returns, good ESG exposure, but not exceptional value vs traditional bonds.

Conclusion

ML attribution reveals green bond premiums reflect supply/demand and selection bias, not fundamental environmental value.