April 2001 Revisions to New Residential Sales

Effective with the April 2001 data release, on May 24, 2001, the Census Bureau made some changes to the methodology used to generate the estimates of New Home Sales.

Summary of Changes

The following items account for most of the changes in our estimates of housing units sold and for sale:

1. Discontinuance of adjustment for construction in areas where building permits are required without a permit being issued. Since the 1960s we have boosted our new home sales data by 3.3% to account for this construction which would not normally be reported in our survey. The housing industry and trade groups believe that such unauthorized construction has virtually ceased, so we have eliminated the 3.3% boost. This was not phased out over time, it was dropped completely in our revised estimates as of January 1999.

2. New grouping of data. To estimate a United States total, we estimate for smaller cells and sum up to the U.S. total. The smaller cells are at geographic level and month permit was issued. Research has shown that reducing the number of cells used in a ratio estimation (which is what we use) will reduce the bias inherent in a ratio estimate.

3. New methods for the calculation of nonresponse and undercoverage adjustment factors (NUAFs). We use these factors to estimate for sales prior to the issuance of a permit, late reports, and corrections to be made. These are being calculated using some revised methodology.

4. The new processing allows more reported data to be included in the revised estimates. Previously we had to allow more time for processing and have an earlier cut-off of additional reported data.

More details on these changes can be found here.

Average and median prices have also been revised to incorporate results from recent methodological changes and a revised outlier procedure. A full write up of the methodology used for medians can be found here.

Effects of Changes on the Estimates

Based on revised data for the years 1999 and 2000, we have estimated the effect of these methodology changes on our estimates. These have been calculated as follows:

The total number of new housing units sold will be about 2.9% lower than previously published. The discontinuance of the 3.3% adjustment on single family houses accounted for the majority of the decline. The new methodology and processing changes resulted in very minor revisions.