EconSouth (Fourth Quarter 2006)
Fed @ Issue
|John Robertson is vice president over the regional group of the Atlanta Fed's research department.||When Things Don't Add Up|
Because the assessment of current economic conditions is an important ingredient in an economic forecast, timely data about economic activity must be obtained from various public and private sources. One indicator of economic performance at the national and regional level is employment, and the most widely used source of timely employment data is the Current Employment Statistics (CES) survey published by the Bureau of Labor Statistics (BLS).
Each month, the BLS surveys some 160,000 businesses and government agencies, which include 400,000 individual work sites nationwide. These surveys measure the number of workers on an establishment's payroll, and the data are also used for other measures such as estimating employment growth by industry and location.
Slow employment growth raises questions
As measured by the CES, employment growth in the United States has remained surprisingly sluggish almost five years after the end of the 2001 recession. For instance, from January 2005 to October 2006, the CES estimates nonfarm employment grew on a year-over-year basis at an average annual rate of 1.5 percent. In contrast, the average annual rate of growth during the boom years of 1996 to 1999 was almost 2.5 percent.
What makes the recent slow employment growth surprising is that other indicators of labor market conditions have been rather strong. The estimate of employment gleaned from a separate BLS survey of individual households, for instance, indicates that the annual rate of job growth has averaged close to 2 percent over the last couple of years. In addition, the unemployment rate has declined steadily since mid-2003 and today is similar to the level seen in 1998—near the height of the information technology boom.
To get a detailed picture of economic conditions, the staff at the Atlanta Fed also consider other sources of information, such as anecdotal reports from regional contacts, including Atlanta Fed directors, and members of the Small Business, Agriculture, and Labor Advisory Council. These various anecdotal reports have reinforced the notion that labor markets are quite tight and that many businesses have been struggling to find qualified workers across a spectrum of industries.
When different sources of information yield conflicting stories we have to dig deeper to see if a potential explanation exists for the discrepancy. Sometimes, anecdotal reports are too narrow to indicate a broad trend. At other times, there can be issues with the data, which brings us back to the recent CES employment numbers.
Annual benchmarking targets errors
The CES employment data are always subject to some sampling error because the CES is a survey, and its sample represents only about one-third of all U.S. nonfarm jobs. Additionally, the BLS must estimate the contribution to employment changes due to business openings and closings. For instance, because a new firm is not immediately included in the CES sample, the BLS estimates the number of jobs associated with business openings based on historical trends. As a consequence, any deviation from the historical pattern will introduce an error into the CES estimate of aggregate employment.
To address these types of errors, each year the BLS realigns the CES employment estimates to payroll employment counts for the month of March. These employment counts are obtained primarily through unemployment insurance (UI) tax records that employers must file with state employment security agencies. Because the UI records cover approximately 97 percent of total nonfarm employment in the United States, they have relatively little sampling error. Unfortunately, the UI data also take quite a while to compile and verify and are currently only available with about a six-month lag—too long to be useful for current analysis work.
On October 6, the BLS announced that, based on a preliminary examination of the UI tax data, its March 2006 estimate of U.S. nonfarm payroll employment would be revised upward by 810,000 jobs, an increase of 0.6 percent. This revision would be the largest in the history of the CES survey. The BLS will publish the detailed results of the CES revision in February 2007.
So, why does this revision matter? For one thing, the revised CES estimate of aggregate job growth between March 2005 and March 2006 will be just over 2 percent, rather than the current estimate of about 1.5 percent. That revision will bring the CES jobs estimate more in line with other labor market indicators.
The revision will have other implications as well. For instance, estimates of average labor productivity will be revised downward because the higher employment will translate into more aggregate hours worked for a given level of output. This revision is significant because the trend in labor productivity is thought to be one determinant of the economy's growth potential.
In addition, the revision will affect post-March 2006 CES employment estimates if the BLS adjusts its estimate of the rate of business formation. Preliminary UI employment data for April–June 2006 are scheduled to be released by the BLS in January 2007, and those numbers will help assess the likelihood of any further revision to the CES employment estimates.
Photo by Flip Chalfant