Foundations of Intrinsic Habit Formation

We provide theoretical foundations for several common (nested) representations of intrinsic linear habit formation. These representations are dynamically consistent and additive, with geometrically decaying coefficients of habit formation. Our axiomatization introduces a revealed preference theory of weaning a decision-maker from her habits using the device of compensation. We characterize linear habit formation in terms of the ability to wean using uniquely determined compensating streams. Moreover, we distinguish between habits that are responsive to weaning and those that are persistent, develop a simple choice-theoretic measure of the rate of habit decay, and demonstrate how to recover the entire sequence of habit formation coefficients from observed choice behavior. We introduce novel monotonicity and separability axioms that are appropriate for time-nonseparable preferences. Our analysis suggests techniques for eliciting dynamic reference points from choice behavior and obtaining discounted utility representations on endogenously generated auxiliary spaces.