This study updates existing literature on consumer/household demand for fiber by examining household purchase dynamics for dietary fiber. It uses a dynamic Tobit model that allows past purchase occasions to affect current purchase decisions for fiber in a framework that captures simultaneously state dependence, unobserved households heterogeneity preferences, and serial correlation caused by a stationary first-order choice process. The model controls for the unobserved heterogeneity by adopting a Gaussian random effects specification. It also captures variations in prices over time and controls for left-censoring. The dynamic model is estimated using the Geweke-Hajivasssiliou-Keane recursive probability simulator and a unique dataset that contains detailed fiber purchase information of households as well as the purchase price, promotion deal, and household demographic information over time.Overall, household purchase decisions are found to be characterized by significant unobserved heterogeneity, statistically significant positive serial correlation, and negative and significant state dependence, implying that lagged purchases have a strong effect on current household decisions such that households purchasing previously at time period t-1 would buy less fiber at time t. Estimation results also reveal that covariates that are not integral determinants of fiber purchases are household participation in the WIC program, the age and presence of children between 13 and 17, not being Hispanic, and the employment level of the female head. Furthermore, the education level of the female head has a negative impact on fiber purchases, whereas use of coupons has the opposite effect.