We study a simple model of economic growth where society’s preferences are a function of consumption per capita and climate quality; and the specification of the climate dynamics is inspired by recent work in climate science. The model is estimated to establish a reference model and we develop a new method that determines the reasonable size of a set of surrounding models which are difficult to distinguish from the reference model. We show that robust agents who deny the effects of climate change on the economy, behave more like agents who believe climate changes are real. This happens because robust non-believers design policies that hedge against their worst case model which does include an anthropogenic effect of their emissions on climate and these changes in climate have negative effects on preferences and productivity.