Greener on the Other Side?

Sammanfattning

Seminar Date: 04.06.2026 Course: BUSN79 - Degree Project in Accounting and Finance Authors: Isak Magnusson & Paula Plinke Advisor/Examiner: Reda Moursli/Diem Nguyen Five Key Words: Green Bonds, Cost of Capital, Greenium, Reputation Effect, Yield to Maturity Purpose: This paper aims to study whether a greenium exists for corporate bonds issued by public firms and the effects of repeated issuance on the greenium and cost of capital. Methodology: The paper applies pooled OLS regressions on matched green and conventional bond pairs, alongside fixed effects panel data models at the firm-year level, controlling for relevant bond and firm characteristics. The dependent variables are YTM, Cost of Debt, and Cost of Equity. The main independent variables are a Green bond indicator, Cumulative Bonds (CB), Cumulative Green Bonds (CGB) and Cumulative Convectional Bonds (CCB). Theoretical perspectives: Efficient Market Hypothesis, Information Asymmetry, Signalling, Investor Clientele, Reputation and CSR Theories. Empirical foundation: The paper applies a bond-level data set consisting of 493 green bonds and 2,207 conventional bonds issued by public firms in the European Union between 2015 and 2025. Based on this sample, 150 matched bond pairs are constructed, alongside a firm-year panel data set consisting of 1,029 firm-year observations across 290 unique firms. Conclusions: The findings reveal no significant greenium in the EU corporate bond market. Additionally, repeated green bond issuance exerts no differential pricing effect compared to repeated conventional bond issuance. However, repeated green bond issuance is associated with a reduction in firms’ Cost of Equity of 6.4 bps per additional green issuance, suggesting that green bond programmes may generate broader reputational benefits. This effect is particularly pronounced among non-financial issuers. In contrast, the baseline Cost of Debt finding is sign inverted relative to the Cost of Equity and its hypothesis, but loses significance across alternative specifications. Thus, no inference is drawn from this finding.