Section on Government Statistics – JSM 2012
Analysis of JOLTS Research Estimates by Size of Firm October 2012
1
Katherine Bauer Klemmer 1
U.S. Bureau of Labor Statistics, 2 Massachusetts Ave. NE, Washington DC 20212
Abstract
The Job Openings and Labor Turnover Survey (JOLTS) program of the Bureau of Labor
Statistics recently produced firm size estimates for research purposes. For each of the 3
firm size classes, data series were produced at the total private level for job openings,
hires, total separations, quits, layoffs and discharges, and other separations from
December 2000 through December 2011. For this paper, firm size classes are compared
to each other, to national level JOLTS data, Business Employment Dynamics net
employment growth, and to Current Employment Statistics by firm size. Differences in
the movements of the time series before, during, and after the most recent recession are
analyzed with respect to size of firm. The impact of the economic cycle on larger versus
smaller firms is discussed as are other factors that potentially contribute to the time series
behavior of the different size classes.
Key Words: firm size class, research data, recession
1. Introduction
The Job Openings and Labor Turnover Survey (JOLTS) program of the Bureau of Labor
Statistics measures job openings, hires, and separations on a monthly basis by industry
and geographic region. 2 The JOLTS program gauges labor demand by collecting data
monthly from a sample of approximately 16,400 nonfarm business establishments.
In 2010, the JOLTS program began producing a research data series by establishment size
for research purposes. An establishment is defined as an economic unit, such as a factory,
mine, store, or office that produces goods or services. Generally, an establishment is at a
single location and is engaged predominantly in one type of economic activity. For each
of six establishment size classes, research data series were developed for job openings,
hires, total separations, quits, layoffs and discharges, and other separations from
December 2000 forward. More recently, the JOLTS program began development of firm
size data. A firm can be defined as an aggregation of establishments under common
ownership. (Okolie 2004, 3-12) A firm can consist of one to many establishments.
While there are uses for both establishment size class data and firm size class data, there
are several factors which the JOLTS program took into account when deciding which
type of size class data to produce. These include sample size considerations (the JOLTS
sample consists of 16,400 establishments), methodological considerations (including the
difficulty in benchmarking the establishment size class data), and size class placement
considerations. (Butani et al. 2012) The JOLTS sampling frame is derived from a
universe of over 9 million establishments of the Quarterly Census of Employment and
1
The views exp ressed are those of the author and do not represent official positions of BLS.
The JOLTS program fo llo ws the North American Industrial Classificat ion System and the term
“industry” can refer to a supersector, sector, or subsector, depending on the context. Census region
descriptions can be viewed at: http://www.bls.gov/eag/.
2
1903
Section on Government Statistics – JSM 2012
Wages, the Unemployment Compensation for Federal Employees, and the Federal
Railroad Administration. Establishments are sampled from nonagricultural industries, all
50 states and the District of Columbia, private sector and government, and all
establishment size classes. It was determined that with a sample size of 16,400
establishments, it would be feasible to produce three firm size classes. Firm size classes
are categorized based on employment from 1-49, 50-499, and 500+. With the firm size
class data, it is also possible to improve over the methodology used for the production of
establishment size class data. The firm size class data can be estimated with techniques
similar to the estimation methodology used for published JOLTS estimates including
independent population controls and alignment at the industry level. Finally, in
establishment size class data, small establishments belonging to large firms are classified
with smaller establishment size classes. In firm size class data, small establishments
belonging to larger firms are classified based on the firm size class. An establishment
within a large firm will be classified as part of the large firm size class.
This paper discusses the initial results from the development of JOLTS firm size data.
Reasonableness of the firm size class data is analyzed by comparing the summed firm
size class data to published JOLTS data at the total private level. Firm size class data are
also compared to Business Employment Dynamics (BED) data, Current Employment
Statistics (CES) firm size class data, and the three size classes are compared to each
other. Differences in the movements of the firm size class time series of the economic
cycle are analyzed with respect to size of the firm. The impact of the economic cycle on
larger versus smaller firms is discussed as are other factors that potentially contribute to
the time series behavior of the different firm size classes.
2. JOLTS Data Elements Defined
Firm size class data were calculated for private sector job openings, hires, and total
separations including quits, layoffs and discharges, and other separations. The firm size
class time series were produced monthly for the period of December 2000 to December
2011. Job openings are collected as of the last business day of the month (stock measure).
A job is considered open if a position exists and work is available, the employer is
actively recruiting from outside the establishment, and the job could begin within 30
days. Hires are collected for the entire month (flow measure). Hires include new hires,
rehires, and recalls from layoffs lasting more than 7 days. Total separations are collected
for the entire month (flow measure) and include quits, layoffs and discharges, and other
separations. Quits are voluntary separations, layoffs and discharges are involuntary
separations, and other separations consist of separations that do not fit either of the first
two categories such as retirements and deaths. JOLTS also collects, but does not publish
employment levels by establishment. Employment includes all persons on the payroll of a
sample unit who worked or received pay for the pay period that includes the 12th of the
month. It includes full- and part-time employees, permanent, short term and seasonal
employees, salaried and hourly workers, and employees on paid vacation or other paid
leave.
3. JOLTS Firm Size Estimates Compare d to JOLTS Summe d Published
Estimates
The initial step in analyzing the accuracy of the new firm size class data series was to
compare the summed firm size class data to published JOLTS data at the total private
1904
Section on Government Statistics – JSM 2012
level. Data from the three firm size classes were summed and compared to top level
published JOLTS estimates. While there are some variations between the summed firm
size classes and the published data, these are minor. Summed firm size class data behaved
similarly to the JOLTS published data throughout the economic cycle. Job openings,
hires, and quits all showed downturns prior to the onset of the most recent recession.
Layoffs and discharges began to climb after the beginning of the recession. This behavior
mirrors the published total private times series. Chart 1 shows the comparison graphically
between the two data series.3
Chart 1. Summed firm size class data, seasonally adjusted, compared to published
JOLTS estimates, seasonally adjusted, by data element
Summed Firm Size Class vs. Published Job Openings
Summed Firm Size Class vs. Published Hires
5000
6000
5500
4000
5000
3500
Hires
Job Openings
4500
3000
2500
4500
4000
2000
1500
3500
1000
3000
Summed Firm Job Openings
Published Job Openings
Summed Firm Hires
Summed Firm Size Class vs. Published Layoffs &
Discharges
Summed Firm Size Class vs. Published Quits
3000
3500
2500
Layoffs & Discharges
4000
Quits
3000
2500
2000
1500
2000
1500
1000
500
0
1000
Summed Firm Quits
Published Quits
Summed Firm Layoffs & Discharges
Summed Firm Size Class vs. Published Other Separations
Published Layoffs & Discharges
Summed Firm Size Class vs. Published Total Separations
500
6000
450
400
Total Separations
Other Separations
Published Hires
350
300
250
200
5500
5000
4500
4000
150
3500
100
3000
Summed Firm Other Separations
Published Other Separations
Summed Firm Total Separations
Published Total Separations
Source: U.S. Bureau of Labor Statistics
4. JOLTS Firm Size Estimates Compare d to BED Firm Size Data
JOLTS firm size class data were also compared to Business Employment Dynamics
(BED) firm size class data. The BED program produces statistics generated from the
Quarterly Census of Employment and Wages. Employment changes are tracked at the
3
For Charts 1, 3, and 4, shaded areas denote recessions as determined by the National Bu reau of
Economic Research.
1905
Section on Government Statistics – JSM 2012
establishment level and then aggregated to produce gross job gains from opening and
expanding establishments and gross job losses from closing and contracting
establishments by industry and firm size class.
BED employment gains and losses data are complementary to the JOLTS hires and
separations data. Net gains can be attributed to declining layoffs or increasing hires while
net losses can be attributed to continued separations or lack of hiring. (Bruyere,
Podgornik, and Spletzer 2011, 16-29) This makes the universe-based BED data a primary
source of comparison for the new JOLTS firm size class estimates.4
In an original unpublished comparison of establishment size class data by James R.
Spletzer, BED microdata were tabulated by establishment size and compared to the
JOLTS establishment size class data aggregated to a quarterly basis. (Spletzer 2010, 1-11)
His findings indicated that the smallest JOLTS establishment size class data had too few
hires and separations. While this finding is being examined, it highlights another area of
weakness in the experimental establishment size class data.
BED data is published by firm size class, so with the availability of JOLTS firm size class
estimates, it is no longer necessary to tabulate BED microdata by establishment size. For
comparison purposes, BED firm size classes were combined to correspond with the three
JOLTS firm size classes. JOLTS firm size class data were aggregated by quarter for
comparability with BED quarterly data. While there are definitional differences between
BED data and JOLTS data including the sizing technique used and the fact that BED is a
universe and JOLTS is sample-based, the comparison is still useful for analysis purposes.
To compare net employment growth of the two series using firm size class estimates,
JOLTS net employment change was calculated by subtracting separations from hires and
BED net employment change was calculated by subtracting BED losses from BED gains.
Correlations were computed for the net employment growth series. The results show that
the smallest firms’ net employment growth series have the lowest correlation at .90 while
the correlation for midsize firms was .93 and for large firms was .91. The correlation for
net employment growth for all firm size classes summed was .96. The correlation results
indicate similar patterns over the two time series.
4
For an overview of the Business Emp loy ment Dynamics program, v isit the BLS website at
http://www.bls.gov/bdm/bdmover.ht m.
1906
Section on Government Statistics – JSM 2012
Chart 2. Net Employment Growth Comparison
Net Employment Growth, All Firm Sizes
1500
1000
500
0
-500
-1000
-1500
-2000
-2500
-3000
corr=.96
JOLTS Net Employment Growth
Net Employment Growth, Firm Size 1: 1-49 Employees
400
200
0
-200
-400
-600
-800
-1000
-1200
BED Net Employment Growth
JOLTS Net Employment Growth
Net Employment Growth, Firm Size 2: 50-499 employees
400
0
-200
-400
-600
-800
corr=.93
-1000
JOLTS Net Employment Growth
BED Net Employment Growth
Net Employment Growth, Firm Size 3: 500+ employees
600
400
200
0
-200
-400
-600
-800
-1000
200
corr=.90
BED Net Employment Growth
corr=.91
JOLTS Net Employment Growth
BED Net Employment Growth
Source: U.S. Bureau of Labor Statistics
5. Comparison of Firm Size Classes 1, 2, & 3
During Alan Krueger’s testimony before the Joint Economic Committee in May 2010, he
used JOLTS establishment size class research data to analyze employment trends in the
U.S. labor market with regards to small, medium, and large businesses by aggregating the
establishment size classes into three categories: less that 50, 50-249, and 250 and greater
employment. (U.S. Senate 2010) Krueger stated that “To better understand the dynamics
behind the dramatic loss in employment that we have experienced in the past two years,
we can examine data on job openings, hires, and separations.” He noted that while job
openings started to fall in early 2007, the job openings level for the largest establishments
experienced its greatest drop at the onset of the financial crisis that started in late 2008.
According to Krueger, the divergence in hiring levels between large and small
establishments was also affected by the financial crisis in 2008. While smaller
establishments initially reacted with layoffs and business closings, the first response of
larger establishments was to freeze hiring. Krueger states that part of this response was
due to the limited access small businesses had to bank financing compared to larger
firms. Quits and layoffs and discharges from establishments with fewer than 50
employees followed a trend similar to that of the total private level. After the financial
crisis, layoffs and discharges for total private establishments as well as for those with
fewer than 50 employees reached significant highs during the first half of 2009 and then
declined steadily through 2010.
Similar to the analysis done by Krueger using establishment size class data, three firm
size class time series were compared to each other by data element. During the most
recent recession, job openings for large firms showed the largest drop during 2008. None
of the firm size classes have recovered pre-recession levels of job openings, however,
post-recession recovery in job openings is most prevalent for large firms.
1907
Section on Government Statistics – JSM 2012
Midsize and large firms cut back on hiring more so than smaller firms. Layoffs and
discharges for small and midsize firms increased over the course of the recession while
quits fell for all three firm size classes. During the recession, job openings, hires, and
quits dropped while layoffs and discharges increased. Following the recession, large
firms appear to make a quicker recovery overall. (Krueger and Charnes 2011, 16-24)
Chart 3. Comparison of Firm Size Class 1, 2, and 3 Data, seasonally adjusted
Total Private Hires, Firm Sizes 1, 2, & 3
2600
2400
2200
2000
1800
1600
1400
1200
1000
800
600
400
200
Hires
Job Openings
Total Private Job Openings, Firm Sizes 1, 2, & 3
Job Openings, Sz1
Job Openings, Sz2
2600
2400
2200
2000
1800
1600
1400
1200
1000
800
600
400
200
Job Openings, Sz3
Hires, Sz1
1200
1000
800
600
400
200
0
Layoffs & Discharges, Sz1
Quits, Sz1
Quits, Sz2
Quits, Sz3
Total Private Total Separations, Firm Sizes 1, 2, & 3
Total Separations
400
350
300
250
200
150
100
50
0
Other Separations, Sz1
Other Separations, Sz2
Layoffs & Discharges, Sz2
Layoffs & Discharges, Sz3
Total Privagte Other Separations, Firm Sizes 1, 2, & 3
Other Separations
Hires, Sz3
Total Private Layoffs and Discharges, Firm Sizes 1, 2, & 3
1800
1600
1400
1200
1000
800
600
400
200
0
Layoffs & Discharges
Quits
Total Private Quits, Firm Sizes 1, 2, & 3
Hires, Sz2
Other Separations, Sz3
2800
2600
2400
2200
2000
1800
1600
1400
1200
1000
800
600
400
Total Separations, Sz1
Total Separations, Sz2
Total Separations, Sz3
Source: U.S. Bureau of Labor Statistics
6. JOLTS Firm Size Estimates Compare d to CES Firm Size Estimates
CES total private data is frequently used for comparisons to the published JOLTS total
private data. The CES program produces monthly estimates of nonagricultural
employment and has a sample size of approximately 140,000 businesses covering
approximately 440,000 individual worksites. Because the CES program also recently
published firm size class research series, it is possible to compare those data to the
1908
Section on Government Statistics – JSM 2012
JOLTS firm size class research series.5 The JOLTS firm size class data appear to be
consistent with the CES firm size class trends at the total private level. See Chart 4.
The summed firm size classes very closely mirror the published data of the two series.
For large firms, job openings level off in advance of the recession and in advance of the
CES large firm employment level data. For midsize firms, job openings started to decline
at the beginning of the recession and employment declined soon after. Job openings and
employment levels recovered more rapidly post-recession for large firms than for the
midsize and small firms.
Hires in all classes leveled off and began their declines prior to December 2007 even
while employment at the firm size class levels continued to grow. Total separations levels
for all three firm size classes exceeded hires while employment declined during the
recession.
Chart 4. JOLTS Firm Size Classes 1-3 Compared to CES Firm Size Classes 1-3, Total
Private Level, Seasonally Adjusted
Firm Size 1: JOLTS Total Separations & Hires vs. CES
Employment
32000
31000
30000
29000
28000
27000
26000
Job Openings Sz1
2200
32000
1800
31000
1600
1000
28000
800
27000
600
26000
Hires Sz1
31000
29000
28000
27000
26000
25000
24000
CES Emplyoment Sz2
2200
CES Employment Sz1
Firm Size 2: JOLTS Total Separations & Hires vs. CES
Employment
32000
2000
31000
1800
30000
1600
29000
1400
28000
1200
1000
27000
800
26000
600
25000
Hires Sz2
5
Total Separations Sz1
Total Separations Sz2
CES Employment Sz2
For mo re informat ion about CES firm size data see “Experimental Size Class Emp loy ment,
Hours, and Earn ings Series fro m the Current Employ ment Statistics Survey” at
http://stats.bls.gov/ces/cessizeclass.htm.
1909
CES Employment
30000
JOLTS Total Separations & Hires
32000
CES Employment
JOLTS Job Openings
29000
1200
Firm Size 2: JOLTS Job Openings vs. CES Employment
Job Openings Sz2
30000
1400
CES Employment Sz1
2000
1800
1600
1400
1200
1000
800
600
400
200
0
33000
2000
CES Employment
JOLTS Total Separations & Hires
33000
CES Employment
JOLTS Job Openings
Firm Size 1: JOLTS Job Openings vs. CES Employment
2000
1800
1600
1400
1200
1000
800
600
400
200
0
Section on Government Statistics – JSM 2012
116000
114000
5000
112000
4000
110000
108000
106000
104000
102000
3000
2000
1000
Job Openings - All
JOLTS Total Separations & Hires
CES Employment - All
JOLTS Total Separations & Hires
6000
CES Employment
JOLTS Job Openings
All Firm Sizes: JOLTS Job Openings vs. CES Employment
118000
53000
51000
49000
47000
45000
Hires Sz3
CES Employment Sz3
7000
Firm Size 3: JOLTS Total Separations & Hires vs. CES
Employment
7000
Total Separations Sz3
CES Employment Sz3
All Firm Sizes: JOLTS Total Separations & Hires vs. CES
Employment
118000
116000
6000
114000
5000
112000
4000
110000
108000
3000
106000
2000
CES Employment
Job Openings Sz3
3200
3000
2800
2600
2400
2200
2000
1800
1600
1400
1200
1000
800
CES Employment
54000
53000
52000
51000
50000
49000
48000
47000
46000
45000
CES Employment
JOLTS Job Openings
Firm Size 3: JOLTS Job Openings vs. CES Employment
3200
3000
2800
2600
2400
2200
2000
1800
1600
1400
1200
1000
800
104000
1000
102000
Hires - All
Total Separations - All
CES Employment - All
Source: U.S. Bureau of Labor Statistics
Subtracting CES previous month employment from current month employment will
result in the net change in employment between periods. This value can be compared to
JOLTS hires minus separations and over time these values should be similar. The JOLTS
estimation process uses the CES employment change to align differences in the JOLTS
hires minus separations data. The correlation between the two summed firm size class
series is .97.
Chart 5. JOLTS and CES Net Employment Growth Comparison, Total Private Level
JOLTS Hires - Total Separations vs. CES net employment change,
All Firm Size Classes
600
400
200
0
-200
-400
-600
-800
-1000
CES Employment All - Net Change
JOLTS All Hires minus Total Separations Net Change
Source: U.S. Bureau of Labor Statistics
1910
Section on Government Statistics – JSM 2012
7. Future Firm Size Research
In the future, it may be possible to publish firm size class data as part of the JOLTS
regular production process but this will depend upon available internal resources and
demand for the firm size data. While the JOLTS program currently plans to provide data
at the total private level only, the reliability of industry level firm size class data is being
investigated for possible future release.
8. References
1. Bruyere, Caryn N., Guy L. Podgornik, and James R. Spletzer. 2011.
“Employment dynamics over the last decade.” Monthly Labor Review August.
Accessed May 3, 2012. http://stats.bls.gov/opub/mlr/2011/08/art2full.pdf .
2. Butani, Shail, Mark Crankshaw, Darrel Greene, and Vinod Kapani. 2012.
Development of JOLTS Firm Size Estimation. Forthcoming in JSM Proceedings,
Alexandria, VA: American Statistical Association.
3. Krueger, Alan B. and Sarah Charnes. 2011. “JOLTS as a timely source of data by
establishment size” Monthly Labor Review May. Accessed May 3, 2012.
http://www.bls.gov/opub/mlr/2011/05/art2full.pdf.
4. Okolie, Cordelia. 2004. “Why size class methodology matters in analysis of net
and gross job flows.” Monthly Labor Review July.
5. Spletzer, James R. 2010. “JOLTS-BED Size Class Comparison.” Internal report,
Bureau of Labor Statistics.
6. U. S. Senate. Joint Economic Committee. 2010. Avoiding Another Lost Decade:
How to Promote Job Creation. Hearing of Written Statement by Alan B. Krueger,
Assistant Secretary for Economic Policy and Chief Economist, U.S. Department
of the Treasury before the Joint Economic Committee, May 5. Accessed: June
21, 2011. http://jec.senate.gov/public//?a=Files.Serve&File_id=6f298a71-cac844fa-95cb-7a47fcae63ee.
1911