In 1983, the Bureau of Labor Statistics converted to a rentalequivalence measure for homeowners’ costs in order to remove theinvestment aspect from the Consumer Price Index for All Urban Consumers.This note provides an estimate of the effect which this change wouldhave had on the CPI-U during 1967-82, when the prices of houses rosesharply. Under the new measure of homeowners’ costs, the index wouldhave increased by 165 percent. By contrast, the official CPI-U climbedby almost 188 percent during the 16-year period. In essence, the change converted the homeownership component from amethod that included investment as well as consumption elements to aflow-of-services approach that measures only the cost of shelterservices consumed by homeowners. The flow-of-services approach uses atechnique that estimates the change in the cost of renting housingservices equivalent to those provided by owner-occupied homes. The change in the index followed many years of recommendations andresearch by BLS staff and by other government, academic, business, andlabor economists and statisticians.
The Bureau conducted extensiveresearch on measuring shelter cost for homeowners as part of the programthat led to the comprehensive revision of the CPI in 1978. From theseefforts came a BLS staff proposal to change the treatment of sheltercosts in the CPI to a flow-of-services approach. This proposal waswidely reviewed and discussed. However, difficulties in developing aworkable flow-of-services measures and the diversity of views held byvarious advisory groups led to the decision not to change thecomponent’s concept at that time.
In 1980, the Bureau introduced five experimental measures (known asthe CPI-U-X1 through the CPI-U-X5) to demonstrate the effect differenthomeownership concepts and techniques could have on the All Items CPI.The experimental CPI-U-X1, which used a rent substitution technique, isthe direct (although approximate) antecedent of the method the Bureauhas now adopted. The development of the CPI-U-X1 increased confidencein the workability and credibility of a rental equivalence measure. Limitations of the CPI-U-X1 Because the CPI-U-X1 was computed outside the CPI production systemby restructuring indexes of major national price groups, it lacked theprecision and detail and, more importantly, the proper local areaweighting of the official CPI.
Further, the five homeownership items ofthe old method–home purchase, contracted mortgage interest cost,property taxes, property insurance, and maintenance and repairs–weresimply replaced in the CPI-U-X1 by a single, new homeownership item.The weight for this new item was computed using a rather imprecise,short-cut technique from homeowners’ estimates of what their homeswould rent for, as reported in the 1972-73 Consumer Expenditure Survey.Finally, the price movement used for this item was the price movement ofthe U.S. residential rent index. BLS addressed these limitations of the CPI-U-X1 for the 1983conversion to rental equivalence. First, the new index was computedfrom local area item-strata cost weights. This process provides theprecision, complete item and geographic detail, and proper geographicweighting associated with the official CPI.
Second, new expenditureweights were calculated by means of the complex statistical estimatingprocedures used throughout the CPI. As described below, weights werecalculated for a new primary homeownership item, homeowners’equivalent rent, and recalculated for household insurance, maintenanceand repairs, and house appliances. Third, a new set of statisticalweights for the sample of rental units was calculated so that its pricemovements could represent owner-occupied housing units as well asrenter-occupied units. Finally, the rent sample was augmented, in partby adding sample units in areas with high owner occupancy.
This madethe sample more efficient for measuring changes in owners’equivalent rent, and the increased sample size reduced the expectederrors of both the renters’ data and the owners’ estimates ofprice change. To address the question of how different the CPI-U-X1 would havebeen had the 1983 refinements been used in its calculation, BLSconstructed an enhanced X1. Unfortunately, it was not feasible toreplicate all of the enhancements.
In particular, recalculating a pricechange measure from an augmented and reweighted rent sample was beyondthe scope of this project. Local area price changes forhomeowners’ costs reflecting the reweighting, but not theaugmentation, of the rent sample were available from some test runs ofthe new procedures starting with data for June 1980. It was notpractical to perform calculations for earlier periods. Nor was itpractical to produce the full index with its full calculationalprecision. The enhanced X1 presented here–like the original–wasproduced using a method that averages together indexes with relativeimportances for weights. However, except for the extensive numeric precision and the item and geographic detail of the officialcalculation, the special technique does yield accurate recalculatedindexes. The first step in calculating an enhanced X1 was to compute aU.
S. rental equivalence index for the 1967-82 period. This index usedthe proper weighting among local areas and, from June 1980 forward,rental change measures from a sample weighted to represent homeowners,instead of renters. The local area indexes were averaged together using the local areaweights for owners’ equivalence of rent that were developed for the1983 CPI-U to produce a national rental equivalence index from the baseyear 1967 to 1982. Compared to the rent weights, owners’equivalent rent shifts weight from the larger metropolitan areas to thesmaller. The three largest local areas account for 29.9 percent of therent weight but only 21.7 percent of the owners’ equivalent rentweight, mostly because of the large difference in the New York standardmetropolitan statistical area.
The CPI geographic areas that representsmaller urban areas have 40.7 percent of the rent weight and 49.5percent of the rental equivalence index.
Renting is more common andmore expensive in large areas. As shown in table 1, the differences between the changes in therent and owners’ equivalent rent indexes were surprisingly small.Reweighting the local area, at least, seems to have no discernible effect, although the change over the 16-year period was slightly smallerfor owners’ equivalent rent.
Reweighting the rent sample itselfmay have some effect. The largest difference between the two seriesoccurs in the last period, the only period in which the full effect ofthe sample reweighting is present. The Bureau, however, does not haveenough information about sample reweighting (and none at all aboutsample augmentation) to be able to demonstrate any differenceconclusively. The next step toward enhancement was to combine the new index ofrental equivalence with other CPI series to obtain an all-items index.The original CPI-U-X1 was an aggregate of the seven major groups of theCPI (food and beverages, housing, apparel and upkeep, transportation,medical care, entertainment, and other goods and services) lesshomeownership (which was part of the housing group) plus the rentsubstitution item. The price movement of the rent substitution itemswas taken from the national rent index.
The enhanced X1 combines a newhousing group with the six other major groups. The housing group is anaggregation of eight items: rent, other rental costs, owners’equivalent rent, household insurance, maintenance and repair services,maintenance and repair commodities, fuels and other utilities, andhousehold furnishings and operations. A small adjustment was made to theweight for household appliances to compensate for the cost implicitycounted in the owners’ equivalent of rent. Table 2 shows thecomponents used to construct the CPI-U-X1 and the enhanced X1, as wellas the weight of each. Although there are more components for housing inthe enhanced X1, the 1983 refinement of the item weights had very littleeffect on the distribution of weight among the major groups. Note thatthe weight for housing in the CPI has been reduced from almost 49percent to about 38 percent.
The results of this study are shown in table 3, which compares theenhanced X1 with the CPI-U-X1 and the CPI-U. From 1967 to 1982,differences between the enhanced X1 and the CPI-U-X1 are few andnegligible compared with the differences between either and the officialCPI-U.