There is a huge variety in models for consumption behaviour of multi-person households that take into account the individual preferences of household members and the resulting household decision process. The difference stems from the modelling of the household decision process, which can go from non-cooperative behaviour up to explicit bargaining models. The main objective of this project is to create a toolbox for a robust comparison of the empirical performances of the various models. Typically, the existing methodology uses the differential approach to derive their characterizations. While this approach has many well-known advantages, it is not optimal for my main objective. Therefore I focus on a revealed preference methodology (RP), which avoids all functional specifications. I will derive RP characterizations for all models of household consumption behaviour. A second main objective is to demonstrate the empirical usefulness of my RP toolkit. Therefore I have to (i) develop richer models that integrate inter-temporal issues, separability assumptions,etc. and (ii) deal with empirical issues such as statistical power, measurement error, heterogeneity,etc. Many of these issues have been integrated in the RP characterizations of the unitary model. Extending these insights to the non-unitary context forms a second type of methodological challenges. In my applications on experimental as well as real-life data, I will (i) demonstrate the versatility of my RP toolkit; (ii) address specific questions in a stylized environment, as such I will, inter alia, contribute to the scarce empirical literature on bargaining models; (iii) illustrate that to tackle welfare related questions, the RP methodology can be complementary to the parametric approach by guiding the researcher in selecting the appropriate non-unitary models and/or functional specifications.