In this paper, a two-equation sample selection model is used to estimate a household demand function for salmon incorporating domestic generic advertising. The twoequation estimation procedure, based on purchase and unit value equations, allows us to handle heavily censored panel data for salmon purchases by Norwegian households and the quality effects simultaneously. Unit values of the aggregated salmon commodity calculated from the observed expenditures and quantities are hypothesized to represent the average quality of the purchased commodity. Advertising effects on both purchases and unit values are investigated. The model also allows us to separate the effects of conditional purchases and purchase probabilities. Results indicate that most (78%) of the advertising effect is through the change of non-purchase occasions to purchase occasions, and that generic salmon advertising induces Norwegian households to spend more money on salmon. However, advertising causes households to select more expensive products rather than increasing their purchased quantities.