Nº 22-64: Different Shades of Green: Estimating the Green Bond Premium using Natural Language Processing

We document the existence of a premium in the green bond market based on the greenness of green bonds. Using BERT, a natural language processing method for textual analysis, we develop a novel measure for bonds’ greenness and document that a 10 percent increase in the bond’s greenness corresponds to a decrease in annualized yield by between 4.86 to 8.71 basis points. In addition to greener bonds enjoying higher premiums, we find evidence that issuing a green bond has positive spillover effects on the pricing of subsequent conventional bonds’ issuance. Overall, our findings are consistent with firms relying on “green" debt instruments to lower capital costs and raise cheaper financing.