U.S. Census Bureau
Manufacturing, Mining, & Construction Statistics Main Page Manufacturing, Mining, & Construction Statistics Main Page Manufacturing, Mining, & Construction Statistics Main Page U.S. Census Bureau Question & Answer Center Business Help Site
MCD Header Image MCD Header Image MCD Header Image MCD Header Image MCD Header Image MCD Header Image MCD Header Image MCD Header Image MCD Header Image
MCD Header Image MCD Header Image MCD Header Image
MCD Header Image MCD Header Image MCD Header Image MCD Header Image MCD Header Image MCD Header Image
MCD Header Image MCD Header Image MCD Header Image MCD Header Image MCD Header Image


SOC Methodology - April 2001: Old versus New Methodology

Component Old Method New Method

All Series

Maximum number of months that permits are tracked for estimation. 84 months. 60 months.
Grouping1 of permit months. 84 individual months. First 12 months are individual. Months 13-18 and 19-60 are collapsed into two groups1.
Adjusting for late reports, undercoverage, and nonresponse. Uses moving averages and fixed ratios to calculate factors. Computes more up-to-date factors. Tests for and adjusts extreme factors. Adjusts for construction activity between 60 and 84 months.
Single family units
Percent adjustment for construction in permit areas without a permit. 3.3% boost. None
Geographic grouping1 Grouped1 by metropolitan and non-metropolitan areas in each of the four Census regions. Grouped1 only by the four Census regions.
Two-unit structures
Geographic grouping1 US Total two-unit structures. Grouped1 by the four Census regions for all structures with two or more units.
Three & four unit structures
Geographic grouping1 US Total 3-4 unit structures. Grouped1 by the four Census regions for all structures with two or more units.
Five or more unit structures
Geographic grouping1 Grouped1 by the four Census regions for all structures with five or more units. Grouped1 by the four Census regions for all structures with two or more units.

1 Grouping is used to adjust the SOC estimates to the Building Permits Series. In the old methods we grouped the data into 1,176 cells. Now we have only 112 cells. Our research has shown that reducing the number of cells reduces the bias of our ratio estimates somewhat. Bias is inherent in ratio estimates, but the high correlation between SOC and Building Permits data yields increases in precision that we believe far outweigh the related small bias.