excellent coverage of multiestablishment firms every five years, but in the years between the economic census, the Census Bureau relies on the COS to update the multiunit segment of the business registers along with information it learns about multiestablishment firms from its other surveys (e.g., Annual Survey of Manufactures4). The limited scope of the sampled firms and the rotation of these sampled firms over time affect the timeliness and coverage of smaller multiestablishment companies in the business register. Several studies have noted that both births and deaths of smaller multiunit establishments are more concentrated in the year prior to the economic census when the register is being prepared for the upcoming economic census.5 One sees this pattern of the clumping of changes in other dimensions of the data as well. McGuckin and Peck (1992) show that industry coding changes for establishments were especially concentrated in the economic census years.

A number of improvements were made to the newest version of the BR, which became fully operational in January 2002: additional data elements were added; seven years of data are now maintained, instead of three, allowing tracking of businesses from one quinquennial census to the next; processing of nonemployer statistics, which previously were not maintained, has been expedited; and industry detail has been brought into concordance with the North American Industry Classification System (NAICS). In addition, in 2005, the IRS began providing quarterly employment data from tax form 941 instead of only for the first quarter. Form 941 includes the EIN, employer-reported wages and other compensation, employment for the pay period, income and social security tax withholdings, and related information. When a new business payroll record is received from the IRS, the Census Bureau adds a business employer record to the BR. Nonemployers cannot be identified as quickly, since personal income tax returns are filed annually rather than quarterly. Form 941 now also includes an identifier for businesses filing final tax returns—useful for capturing business deaths. In July 2004, Census began receiving SS-4 form data directly from the IRS (rather than by way of SSA, as before) on a weekly basis, which allows industry codes to be assigned to new businesses more quickly.6


For noneconomic census years, the Annual Survey of Manufactures provides sample estimates of employment, plant hours, payroll, number of establishments, cost of materials, value of shipments, inventories, and detailed capital expenditures statistics for commercial manufacturing establishments with paid employees (http://factfinder.census.gov).


Jarmin and Miranda (2002) discuss the importance of retiming both small multiunit births and deaths in these data in order to improve the accuracy of the annual birth and death statistics.


Salyers (2004) provides a “progress report” for the BR, including a full description of the expanded use of administrative records, as a more general listing of recent changes and improvements (http://www.stats.gov.cn/english/18roundtable/papers/t20041230_402219768.htm).

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement