Evidence from public goods game experiments holds the promise of informing climate change policies. To fulfill this promise, such evidence needs to demonstrate generalizability to this specific policy context. This paper examines whether and under which conditions behavior in public goods games generalizes to decisions about voluntary climate actions. We observe each participant in two different decision tasks: a real giving task in which contributions are used to directly reduce CO₂ emissions and an abstract public goods game. Through treatment variations in this within-subjects design, we explore two factors that are candidates for affecting generalizability: the structural resemblance of contribution incentives between the tasks and the role of the subject pool, students and non-students. Our findings suggest that cooperation in public goods games is only weakly linked to voluntary climate actions and not in a uniform way. For a standard set of parameters, behavior in both tasks is uncorrelated. Greater structural resemblance of the public goods game with the context of climate change mitigation produces more sizable correlations, especially for student subjects.