In simple economic models, we treat households as insignificant units in the society and assume homogeneity among family members. By doing this, we implicitly assume a unified interest for the household. This is the common preferences model. Although the common preferences model is easy to apply, it is not realistic. A modification would be unified household model, which takes the production and consumption decisions into consideration while treating the household as a single unit. Resources can be efficiently allocated in this case, and one example is Becker’s “rotten kid theorem” suggests that parents can be encouraging their kids to make the right decision for the whole household by appropriate income transfers. But this theorem rests on many assumptions, e.g. the child must choose after a sharing rule is determined, and therefore collapses easily.
For me, the model that makes the most sense is the non-cooperative game model within the household. Each individual calculate their own gains and losses from an activity, and they interact with other family members to maximize their own utilities. Individuals are subject to different threat points (which can be debatable). This setting allows for an inefficient result. Banerjee and Duflo, in their book Poor Economics, mentioned the different yields of women’s and men’s fields and pointed out the potential for further improvement of household welfare, i.e. higher profits. But the household may have deliberately choosen to stick to the less optimal allocation as a respect to the traditional superiority of men over women.
Lee and Bellemare’s paper on mobile phone use in the Philippines (Journal of Development Studies, forthcoming) suggests an interesting relationship between the allocation of mobile phone use within a household and the price information that a household gets. They found that if the phone is in the hands of the farmer or his wife, the price for farm products tend to be higher. Although this analysis is based on a small sample, it is significant in suggesting how policy programs can better improve the welfare of the poor by directing resources to the more efficient users in the household.
A note about fieldwork: In development economics, randomized control trials (RCT) is widely used to ensure the causal relationship between a policy and the goal is measured. One of the difficulties conducting randomized controlled trials in developing countries is that interaction between people happens a lot and is likely to contaminate the results. As suggested by one of my classmates, RCT can be quite difficult in the setting of mobile phone use, because people can borrow each others’ phones. Furthermore, there can always be non-compliers in the experiment, so what we measure is in fact the Local Average Treating Effects, i.e. LATE (see Professor Bellemare’s post).
For a brief introduction to intrahousehold models, please refer to Testing among Models of Intrahousehold Resource Allocation by Cheryl Ross.