People underappreciate how their own behavior and exogenous factors affect their future utility, and thus exaggerate the degree to which their future preferences resemble their current preferences.  We present evidence which demonstrates the prevalence of such projection bias, and develop a formal model that draws out both descriptive and welfare implications of the bias.  The model helps interpret established behavioral anomalies such as the endowment effect, and helps to explain commonly observed suboptimal patterns of behavior such as addiction and excessive pursuit of a high material standard of living.  The model also suggests potentially welfare-improving policies, such as mandatory "cooling-off periods" for certain types of consumer decisions.