The Consumer Expenditure Survey is a nationwide survey of households conducted by the U.S. Bureau of Labor Statistics and U.S. Bureau of Census to find out how Americans spend their money. The current sample design consists of selecting non-contiguous housing units for participation in the survey. This method minimizes variance but can result in high survey costs, particularly in relation to the amount of travel required by data collectors. In this paper we summarize research conducted to determine the feasibility of sampling clusters of contiguous housing units so as to reduce survey costs but still keep variances small.