Measuring the Cost of Shelter for Homeowners: Theoretical and Empirical Considerations

RECENT economic developments have aroused substantial interest in the treatment in the Consumer Price Index (CPI) of the cost of shelter for homeowners.' From December 1977, when the latest version of the CPI was introduced, until December 1980, the all-items CPI increased at an average annual rate of 11.6%, while the homeownership component increased at an average annual rate of 16.2%. Relative to all of the other goods and services in the CPI, the homeownership component has increased by 17.5% over the same time period. If the relative price of homeownership had remained constant the growth rate of the CPI would have been reduced to 10.1%. The question which has been raised is whether the rapid relative increase in the homeownership component, which has had such an important impact on the CPI, truly reflects changes in the cost of shelter. This question is not only important, but difficult, encompassing many subsidiary questions and auxiliary issues. The purposes of this paper are threefold: (1) to outline briefly a conceptual framework for the CPI, which leads to a straightforward specification of what the shelter component of the CPI should measure, (2) to evaluate the theoretical properties of alternative procedures designed to approximate this measurement objective and (3) present empirical evidence on the operational difficulties involved in pursuing a new approach to shelter cost measurement. Two main conclusions are reached. First, on both theoretical and empirical grounds, a "rental equivalence" approach to measuring shelter costs for owner-occupants is preferred. Second, an estimated rental equivalence measure has grown more slowly over (at least) the past six years, than the official CPI homeownership component. Given the way in which the CPI is used to escalate both private and public expenditures, these results demonstrate that the choice of measurement technique has important distributional implications.