The U.S. Census Bureau reported 143,027 high-propensity business applications for October 2021 on November 10, up 0.8% from September after seasonal adjustment. The release also disclosed a revised definition of which applications qualify. For analysts tracking potential new employers, comparing that October number with September’s figure from the previous release would give the wrong direction of change. [S2][S5]
The October 14 report had put September’s high-propensity applications at 145,628. Dividing the new October count by that older September count gives a decline of about 1.8%, rather than the increase of 0.8% reported in November. That calculation deliberately mixes release vintages; it is not an alternative estimate of the monthly trend. The releases do not isolate how much of the difference comes from the definition change. [S2][S5]
High-propensity applications are a subset of business tax-ID applications with characteristics associated with becoming employers. The old definition treated hiring employees, purchasing a business or changing organizational type as qualifying application reasons. The revised version lists hiring employees, but no longer lists the other two reasons. Applications from corporations or with a planned first wages-paid date still qualify. An application that loses one route into the group could therefore qualify through another. [S2][S5]
Industry eligibility changed too. The previous definition included manufacturing under industry codes 31–33 and health care under code 62. The revised definition uses selected manufacturing codes and health-care codes 621 and 623, while adding selected construction, professional-services and educational-services codes. These are changes to the statistical screen, not findings that particular firms hired or dismissed workers. The industry codes belong to the North American Industry Classification System, or NAICS. [S2][S5]
The denominator needs care as well. October’s total business-application count was 432,101, seasonally adjusted. Dividing the high-propensity count by that total gives about 33.1%. This calculated ratio describes the two adjusted application series, not the probability that a business will survive or an observed rate of becoming an employer. Even the total is filtered: Census excludes requests for tax liens, estates and trusts, applications outside the 50 states and District of Columbia, and specified other categories. It is not every request for an Employer Identification Number. [S2]
Census also flagged an annual update involving revisions to past industry classifications, business-formation records and seasonal factors. For a dashboard, use the revised comparison from the new release rather than carrying forward last month’s published number. The two releases do not provide a numerical breakdown of the separate revision effects. [S2][S5]
Nor is the high-propensity count a forecast of how many firms will open. Census separately projected that 31,831 employer businesses would form within four quarters from October’s application cohort, after seasonal adjustment. That is not the number of businesses that started during October. In the November release, the administrative payroll observations used to identify actual employer starts reached only the fourth quarter of 2019. Census said its administrative-data statistics were not subject to sampling error, while warning that projections could be wrong and that recording, coding and other nonsampling errors remained possible. [S2]
Source documents
- Business Formation Statistics, October 2021 (CB21-180)https://www.census.gov/econ/bfs/pdf/historic/bfs_2021m10.pdf
- Business Formation Statistics, September 2021 (CB21-163)https://www.census.gov/econ/bfs/pdf/historic/bfs_2021m09.pdf