This paper uses revealed preference restrictions and nonparametric statistical methods to bound true cost-of-living indices. These are compared to the popular price indices including the type used to calculated the UK RPI. This is used to assess the method of calculating the RPI for substitution bias.