This paper uses the non-parametric techniques developed by Varian (I982, I983) and Afriat (I967, 1973), from the theory of revealed preference - see Samuelson (1947) - to address a number of questions of interest to those who model the determination of consumption, liquid assets and leisure hours.1 The primary assumption underlying much work in this area, whether it deals with highly aggregated categories, such as the consumption of non-durables, or relatively disaggregated categories, such as the consumption of energy products or food, is that demand functions for consumer goods are consistent with the hypothesis that they are derived from the maximisation of a well-behaved utility function subject to a relevant budget constraint. The absence of liquid assets or leisure from such studies involves the implicit assumption of weak separability between consumption goods and these other 'goods'. The parametric approach then assumes a particular functional form for either the utility function, indirect utility function or cost function and estimates the parameters that best describe the data. However, as Varian (i983) points out, when it comes to testing hypotheses in such a framework it is always a joint hypothesis, comprising both the restrictions and the choice of functional form, which is being tested.2 To avoid this problem Varian developed the nonparametric revealed preference approach to provide algebraic conditions which are imposed by maximising behaviour on a set of data, with a finite number of observations. This non-parametric approach is exploited here in several ways. There have been a number of studies using a parametric approach, at a fairly disaggregated level, to the determinants of consumers' expenditure; see, for example, Deaton (I974), Anderson and Blundell (I982) and Ray (I984) for applications using