Does language drive legitimacy? Forest credit issuance and retirement patterns in voluntary carbon markets

Abstract

his paper investigates whether the linguistic content of project descriptions from major voluntary carbon registries can predict credit issuance and retirement outcomes, focusing on forest-based carbon projects. We apply natural language processing techniques including lemmatisation, bigram analysis, semantic families and domain classification. Semantic families and domains are incorporated as explanatory variables in a logistic regression model to assess the predictive power of both lexical variables and structural project characteristics on credit issuance and retirement. Results indicate that text predictors based on semantic domains and lemma families contribute to explaining both credit issuance and retirement probability, though their effects are more modest than those of structural characteristics and somewhat smaller at the retirement stage. Structural characteristics such as registry affiliation, validation body type, and project size consistently dominate explanatory power across the credit lifecycle. These findings raise questions about the informational content of project descriptions and the capacity of existing registry standards to screen for carbon credit quality.