Human behaviour is commonly optimised in economic models of adaptation to climate change. These models assume that people work to maximise profit, subject to financial and technological limitations. In effect, these models simulate adaptive potential. In reality, adaptation falls short of this potential. This shortfall is conceptualised as the adaptation deficit, and it has been causing increasing concern. This study demonstrates the impacts of the ways by which people’s real-world adaptive behaviours depart from those assumed under pure optimisation. These departures, known as adaptation constraints, are formalised as numerical preference functions based on an empirical case study in New Zealand, and they are used to constrain an agent-based model of climate change adaptation. We show that these empirically-specified adaptation constraints reduce profits relative to an optimised specification by roughly one third. This demonstrates that unconstrained economic models are likely to significantly underestimate the costs of adaptation to climate change, the benefits of reducing greenhouse gas emissions, and the residual loss and damage that climate change will cause.