This thesis examines the impact of climate risk on corporate decision-making across three interrelated domains: accounting conservatism, inventory stockpiling, and agency costs. As climate risk emerges as a critical concern in accounting and finance - and as investors and stakeholders place growing emphasis on sustainability - this study advances the existing literature by offering empirical evidence drawn from large international samples.The first empirical chapter investigates how climate risk shapes financial reporting quality, specifically accounting conservatism. Using 20,232 observations across 44 countries from 2008 to 2021, the study finds that climate risk is positively associated with conservatism, with this relationship amplified among firms operating under strong country governance frameworks and those with superior ESG performance. The effect is particularly pronounced for firms in developed economies and environmentally sensitive industries.The second empirical chapter explores the relationship between climate risk and corporate inventory management. Drawing on 42,256 observations across 47 countries over the same period, the study finds that climate risk is positively associated with inventory stockpiling. Product market competition and strong labour unions each reinforce this relationship. Additional analyses reveal that inventory accumulation is more pronounced among firms with elevated agency problems and those in the growth and maturity stages of their life cycle.The third empirical chapter examines how climate risk exposure affects the manager-shareholder relationship. Using 43,729 observations across 33 countries from 2008 to 2020, the study finds that climate risk is negatively associated with agency costs, suggesting that heightened stakeholder scrutiny disciplines managerial behaviour. Analyst following, labour unions, and corporate transparency each moderate this relationship, while managerial entrenchment weakens it.Across all three chapters, findings are robust to a range of endogeneity-mitigation techniques, including quasi-natural experiments, instrumental variables, propensity score matching, and entropy balancing.