ABSTRACT

We are often faced with the problem of having information (variables) in one large-scale high-quality survey that we want to relate to information not collected in that survey but present in others. This paper explores the question of how well we can combine information from multiple survey datasets to make joint inferences. An early example would be the combination of data on domestic work time (collected in time-use surveys) with good information on household income (collected in household budget surveys) to generate a better estimate of household resources (Gershuny and Halpin 1996).