Sample Truncation in FHA Data: Implications for Home Purchase Indexes

Abstract

The home purchase index is the single most important price series in the Consumer Price Index (CPI). In December 1980, for example, the relative importance of home purchase in the CPI for all urban consumers was 10.3 percent. Another component of homeownership cost, contracted mortgage interest, had a weight of 9.8 percent in the CPI. The mortgage cost index is computed as the product of the home purchase index and an index of mortgage interest rates. Thus, a ten percent increase in measured home purchase prices is sufficient to increase the all-items CPI by approximately two percent. The importance of the home purchase index places great value on the accuracy of its measurement. However, the index is one of the most often criticized of all CPI series. The primary objections result from the decision by the Bureau of Labor Statistics (BLS) to base the index on sales data provided by the Federal Housing Administration (FHA). These FHA data have been employed in numerous econometric studies, including analyses of housing demand [14], place-to-place house price indexes [16], and the substitutability of land and capital in the production of housing services [18]. At the same time, the FHA samples are widely recognized to be unrepresentative of the universe of house sales, with possible distorting effects in some applications.' Also, in recent years the home purchase index has differed markedly from other national-level house price indexes published by the U.S. Bureau of the Census and the National Association of Realtors.2 Probably the most critical problem with the FHA data base results from the program's ceilings on the size of insurable mortgages. These ceilings historically have