Future Developments In the Bureau of Labor Statistics Business Employment Dynamics Data By Kristin Fairman and Sheryl Konigsberg Division of Administrative Statistics and Labor Turnover Bureau of Labor Statistics I. Introduction The Business Employment Dynamics (BED) data series highlights the dynamics of the U.S. labor market by revealing the drivers underlying job growth. A fall in jobs lost or an increase in jobs gained can individually create job growth. Job growth as a result of a decline in the numbers of jobs lost has far different policy implications than job growth driven by an increase in job gains. BED data offer policy makers and others insight into the dynamics of job growth and the business cycle. Business dynamics data are currently available at a national level, by industry and by firm size class. Several new BED data series and extensions are under research and development. BED data at the State level are scheduled for release in August 2007. Annual BED data and a study of establishment and firm births and deaths and establishment age will follow. Finally, another ongoing development is the study of entrepreneurship related measures. II. What are Business Employment Dynamics? The BED data demonstrate the theory of creative destruction: the idea that jobs are constantly being created and destroyed by measuring the underlying dynamics of job growth in businesses and employment over time. Job growth is the net result of jobs being both created, or gained, and destroyed, or lost. BED data highlight the dynamics of job growth as well as the dynamics of job gains and job losses. Job gains are the sum of jobs created from existing establishments expanding their workforce and from new establishments opening with new hires. Job losses are the result of jobs destroyed, or lost, from both existing firms contracting their workforces and establishments closing their operations. Job gains, job losses, openings, closings, expansions and contractions are all components of BED data. The Bureau of Labor Statistic s (BLS) Quarterly Census of Employment and Wages (QCEW) is the data source for the BED statistics. The QCEW is a census of all U.S. establishments that have at least one employee and accounts for nearly 98 percent of all U.S. non-farm payrolls. The QCEW starts with administrative data from the businesses that report to the individual State s Unemployment Insurance (UI) system. Covering virtually all businesses, this mandatory reporting of business name, address, industry, employment and wages to the State s UI system is the framework for the nations most timely and complete census of businesses. BLS supplements these data with two additional data collections: the Annual Refilling Survey and the Multiple Worksite Report. First, each quarter, all businesses provide a list of all separate locations allowing BLS to provide accurate employment and wage information for each U.S. county and other areas. Second, BLS updates name, address, physical location, industry and other related information for one-third of all U.S. businesses each year on a threeyear cycle. Also, a careful review of employment and wage levels is conducted each quarter, at both the macro and micro-record level, looking at over-thequarter and over-the-year changes to ensure the changes are the result of economic events as opposed to administrative error. The goal of the review is to improve the quality of the incoming data to yield the highest quality data with timeliness, accuracy and relevance in the U.S. statistical system. As a universe the QCEW data are not subject to sampling or estimation error. In order to create a time-series for each establishment in the QCEW, the establishment data is linked over time to create a continuous history for each establishment. BED data are used to assess the business cycle, the level of labor market volatility and the effect of establishment level employment changes on aggregate employment. A decrease in employment stemming from a decrease in gross job gains has far different business cycle implications than a decrease in employment due to an increase in gross job losses. Many employment statistics only show the decrease in the aggregate employment level but BED data show not only a decline in overall employment but also the root of that decline, helping the business community, economists and policy-makers develop a more complete understanding of the business cycle. Chart 1 highlights gross job gains and gross job losses and Chart 2 highlights the four components that make up the BED data: openings, closings, 469
expansions and contractions for a 10-year period ending in September 2006. Charts 1 and 2 also illustrate the job-less recovery after the 2001 recession (indicated on both Charts by the gray bar) with job growth stemming from declining job losses and not an increase in job gains, which remained below pre-recession levels. The sum of gross job gains and gross job losses illustrates the total number of jobs that are either gained or lost in the economy over the course of one quarter, and may be considered a measure of job churn. BED data demonstrate the tremendous amount of consistent upheaval in a labor market of approximately 114 million jobs with an average of 15 million jobs turning over each quarter. The following sections address the BED data series that are currently available from 1992 onwards: national by industry and by firm size class, and the future developments in BED. Current research and development in BED consist of a near-endless progression of the series with increasing detail at both industry and geographic levels. These series consist of State and county data, annual data, business birth and death statistics, establishment age and survival and entrepreneurship related measures. III. Industry BED Data Overall national job gains remain below prerecession levels, however gross job gains in several industries have returned to pre-recession levels. The construction industry has seen positive job growth with job gains returning to pre-recession levels. This is, in part driven by low home mortgage rates leading families and individuals to increase their demand for new housing (89,000 jobs added in the first three quarters of 2006). As the baby boomers age, there is an increased demand for health care and health care workers spurring job gains in education and healthcare to return to pre-recession levels and positive job growth in recent quarters (291,000 jobs added in the first three quarters of 2006) as shown in Chart 3. Telling a different story, illustrated in Chart 4, the manufacturing sector continues to lose jobs with job gains remaining below pre-recession levels (89,000 jobs lost in the first three quarters of 2006). While the manufacturing sector is not gaining many new jobs, the rate at which the sector is destroying jobs is declining, down from a high of 6.4 percent in the second quarter of 2001 to 4.3 percent in the third quarter of 2006. Gross job gains and losses also remain below pre-recession levels in all other industries. All industries, except manufacturing, exhibit net job gains in recent quarters indicating job destruction fell more than job creation. 470
IV. Firm Size Class BED Data In third quarter 2006, firms with 1 to 4 employees and firms with greater than 1,000 employees shared the largest percentages of gross jobs gained and gross jobs lost. Predictably, the smallest size class, firms with 1 to 4 employees, was responsible for more than half of the jobs gained from openings and jobs lost from closings (59.1 percent and 59.2 percent, respectively in third quarter 2006) while the largest size class, firms with 1,000 plus employees, on average had the lowest share of jobs created from openings and jobs lost from closings (0.1 percent and 0.5 percent, respectively in third quarter 2006). The largest size class had the largest shares of jobs lost from contractions and jobs gained from expansions (19.8 percent and 19.9 percent, respectively in third quarter 2006). Although the largest share of jobs created through openings and the largest share of jobs destroyed from closings are attributed to firms with 1 to 4 employees, these same firms do not contribute the most to job growth. In recent quarters, firms with 1,000 or more employees and firms with 20-49 employees have been the largest contributors to net job growth. Historically, however, no one size class accounts for being the largest contributor to growth. 471
Chart 5. Average quarterly share of net change for firms with 1 to 499 employees and firms with 500 or more employees during the business cycle 90 80 1 to 499 Employess 500 or more Employees 70 60 Share (percent) 50 40 30 20 10 0 Recession 1990/2-1992/1 Expansion 1992/2-2001/1 Recession 2001/2-2003/2 Recovery 2003/3-2005/3 Business cycle period Note: Recession periods in this chart are based on when the BED data exhibited negative net changes and are longer than the NBER (National Bureau of Economic Research) designated date frames for both the 1991 and 2001 recessions. Source: Jessica Helfand, Akbar Sadeghi, and David Talan, Employment dynamics: small and large firms over the business cycle, Monthly Labor Review, March 2007, pp. 39-50. In the March 2007 issue of the Monthly Labor Review, Jessica Helfand and Akbar Sadeghi published an article highlighting some basic findings from the BED size class data. The article, Employment dynamics: small and large firms over the business cycle, indicates that small firms, those with 1 to 499 employees, create about 64 percent of new jobs in the U.S. economy. In addition, the share of growth of small firms is larger than their base share of employment. This growth causes small firms to become large, increasing the employment share of large firms over time. Helfand and Sadeghi also found that firms of different size classes behave differently throughout the phases of the business cycle. During the most recent economic recovery, the contribution of large firms to net job gains appears to come from a fall in the level of gross job losses, rather than increased job creation. The 1991 recession saw the bulk of net job losses occurring in small firms; while large firms generated the majority of job losses during the economic slowdown of 2001 (see Chart 5.) V. Firm Survival Firm survival is of particular interest to anyone starting a business or studying the business cycle and the behavior of new firms. Amy Knaup, in a May 2005 Monthly Labor Review article, Survival and Longevity in the Business Employment Dynamics Data, found that 66 percent of business establishments opening during the second quarter of 1998 were still in existence two years later and 44 percent were still in business four years later. Knaup found no significant differences in these survival rates across industries. A second paper, "Characteristics of Survival: Longevity of Business Establishments in the Business Employment Dynamics Data: Extensions," written by Merissa Piazza and Amy Knaup in December 2006 extends Knaup s original cohort an additional three years. The study found the original cohort had a 31 percent survival rate in the seventh year (see Chart 6). The number of firms exiting the economy declines at a slower rate after the first four years. In addition, survival rates across industries tend to stay consistent over time; in other words, industries with slightly lower survival rates than the average continued to have lower survival rates. 472
Source: Knaup, Amy and Merissa Piazza. Characteristics of Survival: Longevity of Business Establishments in the Business Employment Dynamics Data: Extensions, December 2006. VI. State BED Data One of the more immediate developments is BED data at the State level. This data series will be released in August 2007. In preparation for the release of State BED data, the BED staff conducted an extensive review of employment data for all 50 States and the District of Columbia. The review focused on identifying outliers, where openings, closings, expansions or contractions, were larger or smaller than the expected level in the State. Many of these outliers were caused by data entry errors or missed linkages and were corrected through manual corrections. Other outliers were determined to be a result of the business cycle or administrative errors and were not fixed. In addition to the review conducted by the BED staff, the States also reviewed their individual data. Some of the preliminary findings at the State level show the rates of gross job gains exceeded gross job losses in 21 States in the third quarter of 2006 (see Chart 7). Wyoming, with its strongly seasonal economy, had the highest rate of gross job gains while Alaska had the highest rate of gross job losses in third quarter 2006, the most recent quarter of available data. In third quarter 2006, preliminary data indicates the levels of gross job gains in only 9 States increased over the levels in those States reported in second quarter 2006. Over the same time period, 17 States experienced a decline in the levels of gross job losses. Five States, South Carolina, Kansas, Mississippi, Oklahoma and Nebraska, experienced an increase in gross job gains along with a decline in gross job losses according to early data. However, 30 States experienced both a decline in gross job gains and an increase in gross job losses from second to third quarter 2006. 473
Chart 7. Rates of Gross Job Gains, Gross Job Losses and Net Change by State, June 2006 - September 2006, seasonally adjusted 12.0 1.5 Rates of Gross Job Gains and Gross Job Losses 11.0 10.0 9.0 8.0 7.0 6.0 5.0 0.0-1.5 Rate of Net Change 4.0 Alaska Minnesota Delaware Maine Michigan Montana New Hampshire Vermont Ohio Virginia Kentucky West Virginia Arkansas Maryland Georgia Nevada New Jersey Wisconsin Indiana Connecticut New York California Illinois Missouri National Alabama Colorado Florida Iowa Massachusetts Tennessee North Carolina Pennsylvania District of Hawaii New Mexico Rhode Island Nebraska Washington Mississippi Oklahoma Oregon South Dakota Wyoming Idaho North Dakota Utah Arizona Texas Louisiana Kansas South Carolina -3.0 Gross Job Gains Gross Job Losses Net change VII. Annual BED Data Research on annual BED data is currently under way. Annual BED data compare the employment at the end of one calendar year to employment at the end of the following year, breaking the net change down into its underlying components. Calculating annual BED data is not as simple as summing gross job gains, openings, expansions, gross job losses, contractions and closings for a calendar year. While a data user can sum the net employment change over a year for an annual number, the components can not be summed over the year because this would result in inaccurate counting of the data. For example, an establishment contracts by 50 employees in first and second quarter and then closes its operations, releasing its remaining 100 employees and ending the year out of business. From a quarterly perspective this is seen as a contraction in both first and second quarter of 50 employees and a closing of 100 in the third quarter. By simply adding the quarterly data, we would see a closing of 100 and a contraction of 100, which is not an accurate picture of the underlying annual dynamics. When looking at an annual perspective this establishment should be seen as a closing of 200 since it did exist at the end of the previous year but did not exist at the end of the current year. Therefore, a method was developed to link establishments involved in mergers, acquisitions or other business changes over the year and not just from one quarter to another. A record that both opens and closes during the year is not part of the annual tabulations since it has no impact on employment when comparing the end of one year to the end of the following year. Over calendar year 2005, the most recent year with a full year of data available, preliminary data indicates approximately 27.3 million jobs were created and destroyed. The U.S. labor market gained 14.7 million jobs and lost 12.6 million, resulting in a net increase of 2.1 million jobs to the economy. In addition to the 2.1 million new jobs, 126,000 new establishments were added to the business environment. VIII. Business Establishment Births and Deaths Currently underway is an analysis of establishment births and deaths using BED data. In addition to studying births and deaths, this analysis also allows us to determine an establishment s age. The current BED data does not refer to births and deaths but rather openings and closings. However, births are a 474
subset of openings and deaths are a subset of closings. The BED defines an opening as any establishment with positive employment that did not exist or was reporting zero employment in the one prior quarter while a birth is any establishment with positive employment that did not exist or had zero employment for the prior four quarters. Similarly, BED defines a closing as any establishment that was reporting positive employment in the one quarter but has zero employment or does not exist in the subsequent quarter while a death is any establishment that was reporting positive employment in the one quarter but has zero employment or does not exist in the four subsequent quarters. The difference between openings and births are seasonal re-openings and the difference between closings and deaths are temporary shut downs. In first quarter 2006, there were 219,000 births and 368,000 openings based on preliminary data. In first quarter 2005, there were 185,000 deaths and 348,000 closings based on preliminary data. Sources: Merissa Piazza and Amy Knaup, Establishment Characteristics of Survival: Longevity of Business Establishments in the Business Employment Dynamics Data: Extensions, December 2006. Amy Knaup, Survival and Longevity in the Business Employment Dynamics Data, Monthly Labor Review, May 2005, pp 50-56. Jessica Helfand, Akbar Sadeghi, and David Talan, Employment dynamics: small and large firms over the business cycle, Monthly Labor Review, March 2007, pp. 39-50. IX. Entrepreneurship Measuring entrepreneurship is another area being studied. There are many different definitions of entrepreneurship and BLS is considering which definition best fits the needs of its data users and how BED data can be best used to provide the most relevant information on entrepreneurship. Since the data source used by BED, the QCEW, is a census of almost all U.S. establishments, it provides a rich source from which to study entrepreneurs. The intent of the entrepreneurship study is to examine characteristics of entrepreneurs such as survival rates, age, growth rates and size-of-employment change over time. In addition, BED entrepreneurship data can be combined with other demographic information for a more detailed analysis of the demographic characteristics of entrepreneurs. X. Conclusion Business Employment Dynamics (BED) offers a new perspective on the U.S. labor market by outlining the underlying components of the aggregate employment change in the economy. Since its initial publication of national data in 2003, the range of BED data has consistently been updated and improved in response to user demands with the release of industry data and size class data. In addition to the BED developments underway, other data series such as county and metropolitan statistical area (MSA) data and a study on establishment wages are presently being reviewed for possible publication in the future. 475