The evidence that the same income can lead to different household decisions, depending on who the earner is, has led to an effort to replace the standard household model with the Ôcollective modelÕ, which recognises that a household's decisions depend on the power balance between the husband and the wife. This article recognises that the power balance can, in turn, depend on the decisions made. A new Ôhousehold equilibriumÕ and its dynamics are described and it is shown that there can be multiple equilibria in female labour-supply, and that child labour can decline and rise as the wife's power increases.The unitary model of the household, which had served mainstream economics well for a long time, has in recent times given way to a more fractious view of the household. This has been an outcome of theoretical advances, empirical investigations and anthropological insights.1 It is, for instance, clear that how much say a woman has in the household can vary across households in the same region and with the same total income; and this could depend, for example, on how much income she contributes to the household's total income. This recognition has enormous implications for the design of policy. It means that ÔhowÕ a certain amount of money is injected by government into households can influence the well-being of individuals significantly. Ten dollars given to the male head of the household and the same money given to his wife can have very different implications for not only the amount of tobacco and alcohol purchased by the household but on child labour, education and health (Kanbur and Haddad, 1994). When a series of policy changes in the United Kingdom -see Walker and Zhu (1999) for a description -from 1976 to 1979 caused the household allowance for children to be handed over to the women, instead of men, there was a rise in the expenditure on children's clothings (Lundberg et al., 1997); for related accounts, see Hoddinott and Haddad (1995) and Quisumbing and Maluccio (1999).2 However to go from this broad recognition to the actual design of policy one needs to understand the relation * The author is grateful to George Akerlof, Abhijit Banerjee, Chris Barrett, Alaka Basu, Francois Bourguignon, Nancy Chau, Amanda J. Felkey, Dragan Filipovich, Patrick Francois, Karla Hoff, Ravi Kanbur, Rachel Kranton, Valerie Lechene, Dilip Mookherjee, Omar Robles and Erik Thorbecke for useful discussions and to three anonymous referees for detailed comments and suggestions. The article has also benefited from seminar presentations at