This study examines whether greater discourse on environment-specific political risk (EPR) in earnings conference calls (ECCs) has a favourable effect on firms’ information environment affecting stock price crash risk. Under the risk mitigation hypothesis, EPR discourse lowers crash risk, whereas, under the risk escalation hypothesis, management may exploit information asymmetry leading to greater crash risk. Results from US firm-year observations reveal that the EPR discourse of the current year reduces the crash risk of the following year, thus supporting the risk mitigation hypothesis. Further, this study shows that the effect of EPR discourse on crash risk is more pronounced for politically active firms, potentially due to their proximity to the political decision-making process and better access to associated information. Additionally, the results suggest that the negative association between EPR discourse and crash risk is stronger in the democrat regimes and for the firms with more institutional monitoring and characterised by a higher level of integrity. Random Forest and Extreme Gradient Boosting machine learning algorithms support the effectiveness of EPR discourse in predicting expected crash risk. Earnings forecast dispersion and bid-ask spread appear as potential mechanisms for the underlying association, and the primary results remain robust to potential endogeneity concerns.