Benchmarking Affects Payroll Data

When the September employment situation report is released Friday, one number will loom large, and it won't be the number of new payroll jobs, expected by economists surveyed by Bloomberg to be about 113,000, and it won't be the unemployment rate, expected to be 8.1 percent.

[IMAGE]

Instead, it will be 386,000, which is the number of jobs added to the nation's payrolls, not by employers, but by the ""Bureau of Labor Statistics"":http://www.bls.gov/ in its annual ""benchmarking"" of payroll data. BLS made its ""preliminary"" benchmark announcement last week. It will be officially incorporated into the payroll statistics next February.

According to BLS, total payrolls were 386,000 higher last March than previously reported. Government payrolls were, BLS said, 67,000 lower than had been reported and private sector payrolls 453,000 higher. The revised estimate implies a monthly pace for private payroll gains of about 215,000 in the 12 months through March, rather than the 179,000 per month currently reported.

Though conspiracy theorists may argue otherwise, the revised numbers are not government manipulation of the data but rather are based on more complete jobs data from unemployment insurance records. The unemployment insurance recordsâ€"obtained with the annual filing of unemployment insurance tax formsâ€"not only provide BLS with a more accurate count of payroll jobs, but also provide the BLS with a roadmap to follow in developing its monthly jobs numbers.

BLS develops the payroll estimates from its Current Employment Statistics program, a monthly sample survey of approximately 486,000 individual worksites representing about 160,000 companies reporting on individuals on business payrolls during the week of the month including the 12th calendar day of the month. But since the CES is a sample, it doesn't necessarily cover all the businesses in the country.

That's where the benchmarking comes in. To check its survey results, BLS compares its data with the unemployment insurance returns, which are filed annually and makes the appropriate adjustments â€" either up or down. While the adjustment for 2012 adds 386,000 payroll jobs, the adjustment for 2011 added 162,000, but the adjustment for 2010 subtracted 378,000. The adjustment for 2009 subtracted 902,000.

In addition to being more comprehensive, using the unemployment insurance tax returns has one other advantage: unlike the monthly surveys, which are voluntary, the unemployment tax returns are mandatory.

That participation in the monthly survey is voluntary creates another problem for BLS. If a company fails to respond to the survey in a given month, BLS does not know if the company merely chose not to because of the crush of business or if the company no longer exists.

The unemployment insurance returns serve another function: providing BLS with the universe of businesses from which to draw its CES sample. Start-ups which might not have been required to file unemployment insurance returns because of when they began operations are not in that grouping and therefore not sampled until their second year of operation (assuming they've survived). That could itself lead to an undercount of jobs. BLS compensates statistically by applying a ""birth-death"" model to address the unavoidable lag between a firm opening for business and it appearing on the unemployment insurance roster.

According to BLS, the net of business birth and deaths is relatively small and stable, even during business cycle changes, and BLS research using ten years of unemployment insurance data showed that while birth and death components are large, they are mostly offsetting.

Each September, BLS produces its ""preliminary"" benchmarking to be finalized the following February with data benchmarked to comprehensive counts of employment for the previous March.

Data for March 2012 showed changes of one percent or more to job tallies in the information, construction, mining and logging, and other services industry sectors. Of the 11 major industry sectors, four had fewer jobs than originally reported: government (-67,000), manufacturing (-25,000), professional and business services (-14,000), and education and health services (-1,000). The transportation, trade, and utilities sector had 145,000 more jobs than originally reported, followed by leisure and hospitality (+99,000) construction (+85,000), other services (+54,000), information (+51,000), financial activities (+47,000), and mining and logging (+12,000).

The November and December estimated job counts (released in December and January respectively) are revised due to the combined impact of benchmarking and additional samples as new worksites are rotated into the November survey.