Household beverage consumption is investigated using data from the National Food Stamp Program Survey conducted in the United States. A censored Translog demand system is estimated with the full-information maximum-likelihood procedure. All own-price effects are negative and significant, and whole milk, reduced-fat milk, juice, coffee, and tea are found to be net substitutes for soft drink. Thus, prices provide a partial answer to the declining consumption of milk and rising consumption of soft drink. Nutrition information and dietary beliefs also play important roles, highlighting the importance of an effective nutrition education program directed toward the low-income households. [JEL citation: C34 (Truncated and Censored Models), D12 (Consumer Economics: Empirical), Q18 (Agricultural Policy; Food Policy).] © 2004 Wiley Periodicals, Inc. Agribusiness 20: 309–321, 2004.