We introduce a neureconomic “autopilot” model of habit, based on many studies of animal learning and human habituation. In this approach, there are two systems for valuation– habit and goal-directed. The habitual system recalls the previous choice (which can be dependent on a contextual state), and the reliability of reward of that choice. (Reliability is the absolute value of reward prediction error, so a low value is associated with reliability.) If the reliability is below a threshold the habitual choice is made. Otherwise a goal directed utility-maximizing choice is made. A simple two-choice model is used to show how short-run own-price elasticities can be zero in this model while longer-run elasticities are negative. The theory is quite different than habit formation, as modelled by adjacent complementarity or reference-dependence in other economic theories. In the autopilot model, the structural driver of habit is reward reliability. When this number is persistent, choice is persistent. Persistent behavioral choice is therefore a byproduct of reward reliability and is not the correct preference specification.
