The Quality of New Jobs from the 1990s Through June 2004

CRS Report for Congress
The Quality of New Jobs From the 1990s
Through June 2004
Updated September 22, 2004
Linda Levine
Specialist in Labor Economics
Domestic Social Policy Division
Marc Labonte
Analyst in Macroeconomics
Government and Finance Division


Congressional Research Service ˜ The Library of Congress

The Quality of New Jobs From the 1990s
Through June 2004
Summary
The course of interest in labor market issues thus far in the current decade is
much like the path followed through the mid-1990s. The focus was on people’s job
prospects for at least a year after the 1990-1991 and 2001 recessions ended as the
unemployment rate kept rising and employment kept falling. Once the labor market
rebounded, attention switched to the kind of jobs being created: were the new jobs
good (high paying) or bad (low paying)?
Data are not available on wages paid specifically for newly created jobs. The
quality of job creation was initially approached by examining the industries or
occupations in which employment had grown after dividing them into low paying and
high paying groups. Studies that took an industry approach found most new jobs
added during the 1990s were low wage, while those that took an occupation approach
found most new jobs added were high wage. A limitation of the approaches is the
level of aggregation: although the service-producing sector has been characterized
as low paying, for example, it actually is comprised of industries that pay
comparatively well and comparatively poorly; similarly, industries are staffed by
workers in many different occupations offering a wide range of earnings. Absent
both industry and occupation detail, it cannot be determined whether the jobs being
created are relatively high paid (e.g., physicians), relatively low paid (e.g., nursing
aides), or across-the-board within an industry (e.g., hospitals). To capture these
outcomes, the U.S. Bureau of Labor Statistics developed a matrix of occupation-
industry job categories (e.g., managers in manufacturing) which were then divided
into high, middle, and low earnings employment groups. Based on this approach,
new job growth occurred at opposite ends of the pay spectrum during the 1990s, but
the rate of job creation was greater in the highest compared to the lowest wage group.
The application of this approach to more recent data is problematic because
there are large occupation-industry categories, which account for a large share of net
job growth since 2000, lying near the earnings boundaries that separate lower from
higher paying jobs. Nevertheless, following this approach, Factcheck.org found that
employment rose in high paying jobs and fell in low paying jobs while BusinessWeek
found that employment rose in low paying jobs but rose almost twice as much in high
paying jobs. A problem with these studies is that a job that pays slightly more than
the median wage is treated the same as a job that pays significantly more. One study
that avoided this problem estimated that the wage in expanding industries was 13%
lower than in contracting industries as a share of employment. Most analyses of the
new job quality issue looked at industry data only, and a few, at occupation data only.
As mentioned above, these categories are too broad to give very meaningful results.
Attempts to answer the new job quality question with one “overall” result may
oversimplify. A look at the fastest growing and declining occupation/industry
categories reveals that there is job growth and decline across the earnings spectrum.
In any case, new job quality affects only a small subset of the labor market and may
not offer much illumination on broader labor market issues, such as compensation
growth, unemployment, and earnings inequality. This report will not be updated.



Contents
The Quality of Job Growth During the 1990s............................2
Industry and Occupation Studies..................................2
Results by Industry.........................................3
Results by Occupation......................................4
Occupation-Industry Studies.....................................5
Results ..................................................5
Application of This Approach to More Recent Data...............6
A Review of Recent Studies on New Job Quality.........................7
Industry Studies...............................................8
Occupation Studies...........................................13
Occupation-Industry Studies....................................15
What Can Be Said About New Job Quality?........................18
Uses and Limitations of the Data.....................................20
List of Tables
Table 1. Job Growth and Earnings by Industry (1993-1999)................4
Table 2. Job Growth and Earnings by Occupation (1993-1999).............5
Table 3. Data Used by Recent Studies.................................8
Table 4. Change in Employment by Industry............................9
Table 5. Weekly Earnings by Industry, 2003...........................12
Table 6. Change in Employment by Occupation........................13
Table 7. Weekly Earnings by Occupation, 2003.........................15
Table 8. Employment Change by Occupation-Industry Category,
June 2001-June 2004..........................................16
Table 9. Largest Increases in Employment by Occupation-Industry
Category, June 2003-June 2004..................................19
Table 10. Largest Decreases in Employment by Occupation/Industry
Category, June 2003-June 2004..................................20



The Quality of New Jobs From the 1990s
Through June 2004
For the past few years, the U.S. labor market has performed in what many
consider to be an atypical manner. The longest economic expansion of the post-
World War II period ended in 2001 with a comparatively short, shallow recession.
According to data of the U.S. Bureau of Labor Statistics (BLS), the unemployment
rate rose from a 30-year low in April 2000 of 3.8% to a high of 5.6% in November
2001, the final month of the March-November recession. In the recession year of

2001, employers let go more than 2 million workers in some 21,500 mass layoffs,


and on a net basis, nonfarm payroll jobs fell by 1.6 million or 1.2%. While the
economy then proceeded to pick up steam, the labor market did not: the
unemployment rate continued to climb to a post-recession high of 6.3% in June 2003,
and the net number of jobs and of people employed continued to fall until summer

2003. This pattern rarely occurred following prior recessions.1


Now that the labor market has shown signs of sustained revival, concern has
shifted to questions about the quality of the new jobs being created. That is, some
assert that more of the net employment growth in recent times has been at the lower
end than at the higher end of the pay spectrum while others have made the opposite
claim.2 Just as offshore outsourcing was advanced as a contributing factor to the
sluggishness of the labor market rebound immediately after the 2001 recession, it
also has been offered as perhaps the leading reason for the allegedly low wage
composition of much of the job growth over the past year.3
This report looks into the recent nature of net job growth by earnings level. It
begins by examining the quality of new jobs created during the 1990s and explains
why there are difficulties with applying to recent data a useful methodology
developed during the prior decade. The report then analyzes the empirical studies
that have been conducted to determine the quality of employment growth to date


1 CRS Report RL32047, The “Jobless Recovery” From the 2001 Recession: A Comparison
to Earlier Recoveries and Possible Expanations, by Marc Labonte and Linda Levine.
(Hereafter cited as CRS Report RL32047, The “Jobless Recovery” From the 2001
Recession.)
2 Edmund L. Andrews, “Growth Seems Tilted Toward Low-Wage Jobs,” The New York
Times, Aug. 10, 2004; and David Nicklaus, “Recent Job Growth Is Real — Believe It or
Not,” St. Louis Post-Dispatch, July 23, 2004.
3 Jonathan Weisman and Nell Henderson, “Quality of New Jobs is Focus of Election-Year
Debate,” The Washington Post, June 23, 2004. For information on offshoring see CRS
Report RL32292, Offshoring (a.k.a. Offshore Outsourcing) and Job Insecurity Among U.S.
Workers, by Linda Levine; and CRS Report RL30799, Unemployment Through Layoffs:
What Are the Underlying Reasons?, by Linda Levine.

since the 2001 recession. It concludes with a consideration of the merits of the good
jobs/bad jobs debate.
The Quality of Job Growth During the 1990s
The course of interest in labor market issues thus far in the current decade is
much like the path followed through the mid-1990s. People remained focused on
their job prospects even after the July 1990-March 1991 recession ended as the4
unemployment rate did not show steady improvement in the ensuing 16 months.
The net decline in employment that continued for about a year after the recession’s
end caused the period to be labeled a “jobless recovery.” Once the labor market
began to rebound, attention switched to the kind of jobs being created. Specifically,
were the new jobs good (high paying) or bad (low paying), with the latter often
dubbed “hamburger flipper” jobs in the popular press during the 1992 presidential5
election. The job quality issue continued to garner interest through the 1996
presidential election,6 but then largely dropped from sight during the latter years of
the nation’s longest postwar economic expansion (March 1991-March 2001).
Job quality has many dimensions, among them wages, benefits, and working
conditions. The only one of these variables for which comprehensive data exists over
a long period of time is wages and salaries.
Industry and Occupation Studies
Statistics are not collected on wages paid specifically for newly created jobs.
Labor market observers initially approached the new job quality issue by examining
the industries or occupations in which employment had grown after dividing them
into low paying and high paying groups relative to an earnings standard (e.g., average
or median earnings). Those who took an industry approach utilized the Current
Employment Statistics (CES) survey, which queries 160,000 nonfarm businesses and
government agencies about employment and earnings of wage and salary workers in
production and nonsupervisory jobs.7 Those who took an occupational approach
utilized the Current Population Survey (CPS), which queries 60,000 households in
the civilian noninstitutional population about various labor force variables including
the number of employed persons and their earnings. Although the CES is regarded


4 CRS Report RL32047, The “Jobless Recovery” From the 2001 Recession.
5 Jim Hoagland, “It’s Jobs, Remember?,” The Washington Post, May 13, 1993.
6 In chronological order see for example Beth Belton, “Tale of Two Surveys: Job Quality
Debated,” USA Today, Aug. 8, 1994; George Koretz, “A Trend Toward Quality Jobs?,”
Business Week, Oct. 9, 1995; Jonathan Peterson and Stuart Silverstein, “Quality of Most
New Jobs Better Than Expected,” Los Angeles Times, Mar. 13, 1996; and “Low-Wage
Employment Has Fallen, Report Says,” The Wall Street Journal, Apr. 24, 1996.
7 As the CES is limited to payroll jobs, it does not include the self-employed or unpaid
family workers. The Current Population Survey captures all classes of worker. For more
information see CRS Report RL32387, Self-Employment as a Contributor to Job Growth
and as An Alternative Work Arrangement, by Linda Levine.

as yielding more reliable data in part due to its larger sample size, it does not provide
data by occupation as does the CPS.8
Results by Industry. The U.S. economy has long been characterized by
changes in its industrial structure. The first sector to experience a long-term decline
in employment was agriculture. People left or never entered farming in part due to
plentiful job opportunities that emerged after World War II in manufacturing
industries that were shifting from defense to consumer production (e.g., refrigerators,
washing machines, and televisions). The number of jobs in the goods-producing
sector peaked in 1979, however: overall employment in manufacturing industries
took a downward turn in the 1980s following the same path taken by mining
industries some two decades earlier. The service-producing sector (e.g., wholesale
and retail trade; and finance, insurance, and real estate) has for some time been the
primary generator of jobs, fueled partly by increased consumer demand for a variety
of labor-intensive activities (e.g., the provision of health care and of food away from
home).
While the composition of employment has changed to reflect the reconfigured
industrial structure of the economy, relative wages by industry have remained largely
the same. For example, the average wages paid by retailers and leisure/hospitality
enterprises (e.g., clothing stores and hotels/motels) are below those paid by
construction companies and manufacturers (e.g., commercial builders and
semiconductor producers), which are below the wages paid by professional services
concerns (e.g., lawyers offices and management consultancies). “So changes in the
distribution of employment across industries yield some information about the
growth in good jobs.”9
Of the 20 million payroll jobs created in the 1993-1999 period, the services
industry group (e.g., education and health services) accounted for one of every two
net new jobs.10 Retail trade, another industry within the service-producing sector,
accounted for an additional 17% of job growth. (See Table 1) On average, these
two fast growing industries pay the lowest wages. From this perspective, then, most
of the net job growth during the period occurred in “bad” jobs.


8 The Occupational Employment Statistics (OES) program also provides employment and
wage estimates by occupation and industry, but earnings data were not available before

1996.


9 Alan B. Krueger, “How to Define a Good Job,” The New York Times, Aug. 19, 2004, p.
C2.
10 Council of Economic Advisers (CEA) and U.S. Department of Labor (DOL), 20 Million
Jobs: January 1993-November 1999, Dec. 3, 1999. (Hereafter cited as CEA and DOL, 20
Million Jobs.)

Table 1. Job Growth and Earnings by Industry (1993-1999)
Average Weekly
IndustryShare of TotalNet Job GrowthEarnings of Productionand Nonsupervisory
(%)Workers in 1999 ($)
Goods-producing sector, excluding agriculture
Mining-1737
Construction 9 672
Manufacturing 1 579
Service-producing sector
Transportation and public utilities5607
Wholesale trade5559
Retail trade17264
Finance, insurance, and real estate5529
Services50436
Government8 —
Source: Employment data from Council of Economic Advisers and U.S. Department of Labor, 20
Million Jobs: January 1993 - November 1999; and U.S. Bureau of Labor Statistics, Employment and
Earnings, Jan. 2003.
Note: Based on the 1987 Standard Industrial Classification system. Earnings data not available from
the CES for the government industry group.
Results by Occupation. Just as there have been employment shifts across
industries, so too have there been shifts across occupations. The widening wage gap
between less and more educated workers in the last few decades indicates that
employers have increased their relative demand for persons in occupations typified
by high educational requirements (e.g., a bachelor’s degree).11 Although employment
in these occupations generally has expanded more rapidly than in occupations with
lesser educational requirements, the occupational composition of total employment12
has remained largely unchanged. As with employment by industry and despite the
increased earnings premium paid to occupations employing comparatively highly
educated persons (which occurred to the greatest extent during the 1980s), the
occupational wage hierarchy has remained much the same (e.g., accountants and
registered nurses typically earn more than electricians and truck mechanics who earn
more than office clerks and retail salespersons who typically earn more than child
care providers and food preparation workers).


11 CRS Report 95-1081, Education Matters: Earnings by Highest Year of Schooling
Completed, by Linda Levine.
12 CRS Report 97-764, The Skill (Education) Distribution of Jobs: How Is It Changing?,
by Linda Levine.

Of the 20 million payroll jobs created in the 1993-1999 period, managers and
professionals each accounted for more than 30% of employment growth. (See Table
2) The two occupational groups also made up “almost 70 percent of all college
graduates in the workforce.”13 These fast growing occupations typically pay the
highest wages. From this perspective, then, most of the net job gain during the
period occurred in “good” jobs.
Table 2. Job Growth and Earnings by Occupation (1993-1999)
Median Weekly
IndustryShare of TotalNet Job GrowthEarnings of Full-timeWage and Salary
(%)Workers in 1999 ($)
Manage rial 32.5 792
Professional 31 800
Technicians1618
Sales13523
Clerical-2.5447
Crafts9594
Operator/Laborer 6 429
Service11336
Source: Employment data from Council of Economic Advisers and U.S. Department of Labor, 20
Million Jobs: January 1993 - November 1999; U.S. Bureau of Labor Statistics, Employment and
Earnings, Jan. 2000.
Note: Based on the 1980 Standard Occupational Classification system.
Occupation-Industry Studies
A limitation of the industry and occupation approaches concerns the level of
aggregation. Although the service-producing sector is often characterized as low
paying, for example, it actually is comprised of industry groups that pay
comparatively well (e.g., financial activities) and comparatively poorly (e.g., retail
trade). Similarly, industries are staffed by workers in numerous occupations that
afford a wide range of earnings. Absent occupational detail, it cannot be determined
whether employment is being created in relatively high paid jobs (e.g., physicians),
relatively low paid jobs (e.g., nursing aides), or across-the-board within an industry
(e.g., hospitals).
Results. To capture all these possible outcomes, BLS developed a matrix of
occupation-industry job categories (e.g., managers in wholesale trade). The job
categories were then ranked in descending order based upon median weekly earnings


13 CEA and DOL, 20 Million Jobs.

and divided into three approximately equal size employment groups corresponding
to high, middle, and low earnings.
“As the U.S. economy moved out of the recession of the early 1990s and
employment expanded, job growth in the highest earnings group accelerated” and
was greater than that of the two lower earnings groups over the 1989-1999 period at
27%.14 Almost all managerial and professional occupations were in the highest
paying cluster of jobs, and these highly paid managers and professionals accounted
for a disproportionate share of the cluster’s net job growth. The services industry
increased its employment of managers and professionals to such an extent that the
two occupations in the industry made up two-thirds of the net employment gain
recorded in the highest paying group of jobs.
Employment grew minimally (1%) in the middle earnings cluster. Some notable
employment change did occur in occupation-industry job categories within the
group, however. Blue-collar positions in construction and in transportation/public
utilities greatly expanded, but the opposite was true for blue-collar factory jobs.
Clerical staff at manufacturers fell substantially as well. In contrast, the ranks of
managers in retail establishments grew considerably as did technicians in the services
industry.
Employment in the lowest paid cluster increased by 16%, or at two-fifths the
rate of the highest paid group (27%).
Employment of the lowest earnings group was relatively unaffected by the
recession of the early 1990s. Indeed, through 1993, the rate of employment
growth among low-wage workers actually exceeded that for workers at the upper
end of the earnings spectrum. However, by the mid-1990s, job growth in the15
high earnings group had outpaced growth in the lowest earnings group.
The lowest earnings cluster was mostly comprised of blue-collar, sales, service, and
clerical workers in the retail trade and services industries. The fastest growing of
these occupation-industry categories, which accounted for two-thirds of the low paid
group’s employment gain, were sales and service workers in retail trade and services
industries. Job opportunities for low paid clerical workers varied by industry, with
substantial employment gains in the services industry but cutbacks at retailers.
When employment growth during the 1990s was analyzed simultaneously by
occupation and industry, then, both ends of the pay spectrum recorded the most new
jobs; however, the rate of job creation was much greater in the highest compared to
the lowest wage group.
Application of This Approach to More Recent Data. Unfortunately, the
occupation-industry matrix approach cannot be successfully applied to more recent


14 Randy E. Ilg and Steven E. Haugen, “Earnings and Employment Trends in the 1990s,”
Monthly Labor Review, Mar. 2000, p. 22. (Hereafter cited as Ilg and Haugen, Earnings and
Employment Trends in the 1990s.)
15 Ibid., p. 27.

data because there are large job categories lying near the earnings boundaries that
separate lower from higher paying jobs. Compounding the analytical problem is that
these same job categories have accounted for a substantial share of net employment
growth since 2000. Some illustrations might clarify the matter.16
!Employment in the occupation-industry categories cannot be divided
into approximately equal thirds to correspond to low, medium, and
high paying jobs because the median earnings of professionals in the
education and health services industry lie at the bottom of the high
paying group. But, placement in the highest paid group of the entire
large (13.7 million) occupation-industry category — which
accounted for 1.6 million of the 1.4 million net increase in persons
employed between 2000 and a 12-month moving average ending in
June 2004 — would greatly influence the results.
!The sensitivity of the results to large occupation-industry categories
that lie near earnings boundaries is not obviated by switching from
thirds to halves. If occupation-industry categories are divided in half
based upon those above and below the median earnings in 2003 of
$541, to correspond to low paying and high paying jobs, then there
is a problem with analysis of the construction and extractive
occupations in the construction industry. The median earnings of the
occupation-industry category, which accounted for 850,000 of the
2000-June 2004 employment increase, were $553 in 2003. If the
entire large (6.3 million) occupation-industry category were placed
among the high paying jobs, it would greatly affect the results and
cause half of base period employment to no longer lie below
(above) the earnings boundary.
BLS had been able in its occupation-industry analyses covering the 1990s to
produce robust results, that is, outcomes were similar when employment was ranked
in various base periods of employment. The problem of large occupation-industry
categories falling near earnings boundaries and those same categories accounting for
much of total job growth were not factors then as they are today.
A Review of Recent Studies on New Job Quality
Like the analyses of data from the 1990s, studies have looked at recent changes
in employment by occupation, industry, and by combining occupations and
industries. Table 3 presents selected recent studies which are reviewed below
according to the type of data on which they were based.


16 This information was provided by BLS in telephone conversations with CRS in Aug. and
Sept. 2004.

Table 3. Data Used by Recent Studies
Occupation
DatesOccupationIndustryand Industry
Author of StudyCoveredData OnlyData OnlyData
Joint Economic6/03-6/04x
Committee
Chicago Fed12/03-6/04x
CIBC World Markets10/01-5/04x
Wachovia 5/03-5/04 x
Morgan Stanley6/03-6/04;xx
2/04-6/04
Factcheck.org 6/03/6/04 x
BusinessWeek 6/03-6/04 x
Employment Policy5/03-5/04x
Foundation
Economic Policy3/03-5/04;xx
Institute 6/03-6/04
Industry Studies
Most recent studies have only classified jobs by industry and utilized data from
the CES (establishment) survey. Net employment increased by 1.4 million between
June 2003 and June 2004. As shown in Table 4, the increase was concentrated in
construction and the following service-producing sector industry groups:
professional and business services, education and health services, leisure and
hospitality, and retail trade. Employment in these five industries increased by 1.4
million in that period. Only two industry groups, manufacturing and information,
experienced net job losses in that time. Viewed from the beginning of the last
recession, the picture is a little different. Six industry groups experienced an increase
in employment since March 2001: construction; and in the service-producing sector,
financial activities, education and health services, leisure and hospitality, other
services, and government. The job gains from March 2001 to June 2004 in education
and health services were about two and a half times larger than the next largest
industry. Among the other seven industry groups that saw a net job loss over that
period, manufacturing by far recorded the biggest decrease;17 information was a
distant second. The fact that six out of 14 industry groups experienced a change in
employment over the past year that was in the opposite direction of the change since
the beginning of the recession points to a problem inherent in these studies — the
results will be sensitive to the dates used.


17 For more information on manufacturing employment, see CRS Report RL32350,
Deindustrialization: The Roles of Trade, Productivity, and Recession, by Craig Elwell.

Table 4. Change in Employment by Industry
Change in Employment
(thousands of workers)
June 2003-JuneMarch 2001-June
Indust ry 2004 2004
Goods-producing sector
Natural Resources and Mining16-21
Construction 194 48
Manufacturing -119 -2,536
Service-producing sector
Wholesale Trade30-183
Retail Trade148-297
T r ansportation 54 -218
Utilities5-18
Information -19 -543
Financial Activities48238
Professional and Business Services484-313
Education and Health Services3151,426
Leisure and Hospitality244344
Other Services20219
Public Administration-38586
Source: Created by CRS from BLS data from the CES based on the 2002 North American Industry
Classification system.
Note: Employment peaked in Mar. 2001. Not all studies use the industry groupings shown in this
tab le.
Using data classified by the 14 industry groups in Table 4, a Chicago Federal
Reserve study found that from December 2003 to June 2004,
Those sectors paying above the national average (wage) constitute just under
three-fourths of total employment growth (including five of the seven fastest18


growing industries), despite representing only 65% of total employment.
18 Daniel Aaronson and Sara Christopher, “Employment Growth in Higher-Paying Sectors,”
Chicago Fed Letter 206, Sept. 2004, p. 2. (Hereafter cited as Aaronson and Christopher,
(continued...)

Thus, high paying industry groups accounted for most job growth both on an absolute
basis and as a share of employment. However, different results were obtained when
the data were disaggregated into 84 industries: “... above-average wage sectors added

41% of recent employment gains, while constituting 51% of total employment.”19


The researchers found that much of the difference was because job growth within the
professional and business services group occurred in industries with below average
wages. They found that over the past four business cycles, the quality of new jobs
has been highly cyclical, with new job quality deteriorating in recessions.
Wachovia compared job growth in the 10 industries with above average
earnings to job growth in the four industries with below average earnings. It found
that employment in the high wage industries increased by 1.1 million while
employment in the low wage industries increased by 0.3 million from May 2003 to
May 2004. From these results, one could conclude that most job growth occurred in
high wage industries, but since low wage industries accounted for only about 30%
of total employment, it would be better to compare growth rates of the two to
determine if industries with above average wages grew more rapidly than in
industries with below average wages. By this measure, Wachovia reported that high
wage jobs expanded by 1.2% and low wage jobs expanded by 0.8% in the past year.
This was not the case in the recession or early stages of the recovery, however, when
employment in below average wage industries fell more slowly than in above average
wage industries.20
Morgan Stanley characterized the jobs created from February to June 2004 as
low quality. It estimated that 25% of employment growth occurred in three
industries “at the low end of the job hierarchy” — restaurants, temporary help, and
building services. Another 19% of job gains occurred in nine other low paying
industries (i.e, clothing stores, couriers, hotels, grocery stores, trucking, hospitals,
social work, business support, and personal and laundry services). Thus, Morgan
Stanley estimated that 44% of job creation over a four-month period in 2004 was in
low quality jobs. Alternatively, 29% of employment growth occurred in industries
that Morgan Stanley categorized as high wage — construction (17%) and the
following services industries (12%): legal, architecture and engineering, computer
systems design, consulting, credit intermediation, brokerage and securities firms, and
private education.21
CIBC World Markets concluded that from October 2001 to May 2004,
employment in low paying industries increased by about 1.5% and employment in


18 (...continued)
Employment Growth in Higher-Paying Sectors.)
19 Ibid., p. 2.
20 Wachovia, Stop Whining About the ‘Quality’ of Jobs, Economics Group newsletter, June

4, 2004.


21 Stephen Roach, “America’s Job-Quality Trap,” Morgan Stanley Global Economic Forum,
July 9, 2004. (Hereafter cited as Roach, America’s Job-Quality Trap.) Note: The study
also analyzed the occupational data from the CPS. That part of the study will be reviewed
in the next section.

high paying industries decreased by about 2.5% (high and low paying was not
defined). CIBC World Markets also found that industries gaining jobs had average
earnings 30% lower than industries losing jobs. It went on to note that in 2004 the
rate of job creation in high paying industries accelerated (e.g., natural resources,
construction, and professional services), which CIBC ascribed to “normal cyclical
behavior.”22
One shortcoming of all of these studies is that they do not draw any distinction
among jobs within the high wage and low wage groups. For example, the studies do
not make any distinction between employment growth concentrated in jobs that paid
far above the average wage and growth concentrated in jobs that paid only one dollar
more than the average wage.
The Economic Policy Institute (EPI) utilized a unique methodology that
attempts to avoid this problem. It assigned each industry a weighted share, so that
industries with greater contractions or expansions weigh more heavily in the overall
results. Based on the weighted shares, EPI calculated a composite wage for
expanding categories, and compared it to the composite wage in contracting
categories. The employment data were disaggregated into 20 (private-sector)
industries and three month averages (March to May). From March 2003 to May
2004, EPI estimated that average wages in expanding industries were 13% lower than
in contracting industries.23 The Chicago Federal Reserve study described above
applied a methodology similar to that of EPI, and estimated that from December
2003 to June 2004, earnings in expanding industries were 7% below those in
contracting industries.”24
These EPI and Chicago Fed findings are subject to misunderstanding because
expanding and contracting industries were defined in relative, rather than absolute,
terms. When most people ask “what type of jobs are being created?,” it is likely that
they are thinking of growth in absolute terms — literally, how many thousands of
jobs in any given industry have been added. This is not how the two analyses
defined expanding and contracting industries. Instead, expansion was measured
relative to an industry’s share of total employment. Thus, if an industry experienced
a sizable increase in employment in absolute terms, but not as fast as overall job
growth, it was counted as contracting. There is nothing wrong with this
methodology, but it may not be what the lay reader expected.
A disadvantage of analyses based on industry data is that the industry
classifications utilized are too broad to provide very meaningful results. Each
industry spans jobs from the unskilled to the highly skilled, which can be seen by
looking at an industry’s median wage by occupation. As shown in Table 5, every
industry contained an occupation that paid substantially above the industry’s median
wage and an occupation that paid substantially below its median wage. A one


22 Benjamin Tal, Assessing U.S. Job Quality, CIBC World Markets newsletter, June 21,

2004, p. 3.


23 Employment Policy Institute, Jobs Picture, June 4, 2004.
24 Aaronson and Christopher, Employment Growth in High-Paying Sectors.

standard deviation change shifts every low wage industry except leisure and
hospitality above the overall median wage ($541 in 2003) and all but two high wage
industries (mining and public administration) below the median wage.25
Table 5. Weekly Earnings by Industry, 2003
(in dollars)
By Occupation
MedianStandard
Industry(total = $541)DeviationMinimumMaximum
Goods-producing sector
Agriculture and Related376194314979
Mining 786 295 497 1496
Construction 593 208 312 970
Manufacturing 620 262 386 1125
Service-producing sector
Wholesale Trade630210326943
Retail Trade391166284851
Transportation and Utilities6801914581064
Information 706 261 160 1101
Financial Activities636207300936
Professional and Business617226325986
Services
Education and Health548217112877
Services
Leisure and Hospitality292175123685
Other Services401166266827
Public Administration722154369921
Source: Created by CRS from BLS data from the CPS based on the 2000 Standard Occupational
Classification system and the 2002 North American Industry Classification system.
Note: Summary statistics based on the unweighted data for 11 occupations in each industry.


25 Standard deviation is a common measure of variance that tells how tightly the data points
in a sample are clustered around the mean. With a normal distribution, about two-thirds of
the data points will fall within plus or minus one standard deviation from the mean.

Occupation Studies
Few recent studies relied on occupational data, the leading source of which is
the CPS. In addition to the previously discussed advantages of utilizing CES rather
than CPS data, not all CPS data are seasonally adjusted, so data from different
months within a year cannot be compared.
Job gains by occupation from June 2003 to June 2004 were concentrated in the
professional, service, construction and extraction, and transportation and material
moving occupations. (See Table 6.) Employment in these four occupations
increased by 1.6 million — more than the total net increase in employment. The
same occupation groups along with management, business, and financial operations
dominated job growth from June 2001 to June 2004. Service occupations accounted
for a larger proportion of job gains over three years than over the last year. In the last
year, employment declined in the office and administrative support and production
occupations. Those same occupations and farming, fishing, and forestry lost
employment over the past three years.
Table 6. Change in Employment by Occupation
Change in Employment
(thousands of workers)
June 2003-June 2001-
OccupationJune 2004June 2004
Management, Business, and Financial Operations44630
Professional and Related279849
Service3361,286
Sales and Related174215
Office and Administrative Support-133-422
Farming, Fishing, and Forestry10-59
Construction and Extraction503660
Installation, Maintenance, and Repair170335
Production -434 -1,429
Transportation and Material Moving44358
Source: Created by CRS from BLS data from the CPS based on the 2000 Standard Occupation
Classification system.



The majority staff of the Joint Economic Committee concluded that employment
growth between June 2003 and June 2004 was mostly in well paid occupations based
on their finding that 71.4% of the growth during that time occurred in the
management, business, and financial operations; professional and related;
construction and extraction; and installation, maintenance, and repair occupations.26
The result was reached by dividing the job gain in those four occupations by the net
increase in total employment. If the four occupations’ employment growth were
instead divided by the increase in employment in only growing occupations, the ratio
falls to about 50%.
Morgan Stanley characterized job creation from June 2003 to June 2004 as low
quality. The study found that 81% of employment growth in that period occurred in
occupations which Morgan Stanley characterized as “low-end” — transportation and
material moving; non-professional services; sales; and installation, maintenance, and
repair.27 As shown in Table 7, however, the installation, maintenance, and repair
group paid well above the median wage in 2003.
Unfortunately, the occupation groups in these studies are too broad for their
results to be very meaningful. While occupations might be expected to have less
variance around the median wage than industries, that is not the case. Median
earnings in sales occupations, for example, span from the very low ($215 for sales
workers in the leisure and hospitality industry group) to the very high ($1,496 for
sales workers in the mining industry group). As can be seen in Table 7, every
occupation contained an industry that paid substantially above the occupation’s
median wage and all but one (management) contained an industry that paid
substantially below the median wage. A one standard deviation change shifts five
low paying occupations (sales, transportation, production, office and administration,
and farming, fishing, and forestry) above the overall median wage ($541) and two
high paying occupations (protective service, and construction and extraction) below
the median wage.


26 Joint Economic Committee, Recent Job Growth is Mostly in Well-Paid Occupations, press
release, July 7, 2004.
27 Roach, America’s Job-Quality Trap. Note: This study also analyzed industry data from
the CES, which is reviewed above.

Table 7. Weekly Earnings by Occupation, 2003
(in dollars)
By Industry
O ccupat i on Median St andard
(total = $541)DeviationMinimumMaximum
Manage me nt 932 145 649 1125
Professional 762 195 469 1070
Protective Service588123276738
Service 299 127 160 723
Sales 448 319 215 1496
Office and Administrative47773350618
Support
Farming, Fishing, and342192112737
Forestry
Construction and Extraction577117413802
Installation and Repair653104497793
Production 501 132 311 741
Transportation and Material472127287715
Moving
Source: Created by CRS from BLS data from the CPS based on the 2000 Standard Occupational
Classification system and the 2002 North American Industry Classification system.
Note: Summary statistics based on the data on the 14 industries for each occupation.
Occupation-Industry Studies
Because the 11 industry and 14 occupation groups utilized in the above-
described studies are so broad, crossing the two can break the data into 154
categories with less wage variance. Unfortunately, greater precision comes at the
expense of statistical robustness. Splitting the data into occupation-industry
categories does not result in categories of a uniform size. Category size varies from28
fewer than 1,000 workers to more than 15 million workers. While the average
category contains 932,000 workers (less than 1% of total employment), the standard
deviation is more than twice the mean. When one considers that a category of 1,000
workers is less than 0.001% of total employment and the data come from a survey of
60,000 households, it becomes clear that there are very few observations for the
smaller categories; accordingly, the many smaller categories with small increases or29


decreases cannot be considered statistically significant (different than zero).
28 Calculated by CRS from BLS data.
29 BLS considers earnings data for occupation-industry categories of less than 50,000
(continued...)

Another drawback to the occupation-industry studies is that they, like the
industry studies, are sensitive to the starting and ending dates chosen. For example,
categories that added jobs in the past year may have lost them since the recession
started. These categories may have very different prospects for the future than
categories that have continually added jobs since the recession. As shown in Table
8, in about one-third of all categories, the employment change over the past year was
in the opposite direction of the change over the past four years.
In addition, these studies usually are based on the 154 occupation-industry
matrix that BLS developed and successfully applied to employment and earnings data
from the 1990s. But, as discussed above in the “Application of This Approach to
More Recent Data” section, application of the 154 occupation-industry categories to
more recent data is problematic.
Table 8. Employment Change by Occupation-Industry Category,
June 2001-June 2004
Employment ChangeIncrease, June 2001-June 2004 (%)Decrease, June 2001-June 2004 (%)
Increase, June 2003-June 20042820
Decrease, June 2003-June 20041632
Source: Created by CRS from BLS data from the CPS.
Note: Numbers do not add to 100% because seven occupation-industry groups with less than 1,000
workers are not included. For purposes of this table, zero change is considered a decrease.
Factcheck.org found that from June 2003 to June 2004, employment increased
by 1.2 million in occupation-industry categories paying above median wages (of $541
per week) and decreased by 6,000 in occupation-industry categories paying below
median wages. Higher paying categories that experienced large net job gains
included construction and extraction occupations in construction, professional
occupations in education and health services, and sales occupations in wholesale
trade. Lower paying categories that experienced large net job losses included
production occupations in manufacturing, and office and administrative occupations
in information. Over a longer (3-year) time frame, Factcheck.org estimated that
employment increased by 1.6 million in high wage categories and decreased by

547,000 in low wage categories.


BusinessWeek found that from June 2003 to June 2004, employment increased
by 744,000 in occupation-industry categories paying median wages of at least $559
per week and increased by 408,000 in occupation-industry categories paying median
wages of $553 or less. (No categories had median earnings between $553 and $559,


29 (...continued)
workers to not be statistically robust. See Ilg and Haugen, Earnings and Employment
Trends in the 1990s, p. 32.

according to the analysis.) Although employment in the high wage group represented
48% of total employment, it accounted for more than half of net job growth over the
last year. BusinessWeek reported that high quality employment growth came from
“a variety of sources” (e.g., professionals in wholesale trade and production workers
in mining).30
Although Factcheck.org and BusinessWeek concluded that most net jobs were
high wage ones, the difference in magnitude between their growth estimates
primarily is due to their allocation of one occupation-industry category. The job
category composed of construction and extraction occupations in the construction
industry — which gained 528,000 jobs according to Factcheck.org — was placed in
the above median wage group by Factcheck.org and in the below median wage group
by BusinessWeek. As previously explained in this report, the nearness to the earnings
boundary of large occupation-industry categories that have accounted for a
considerable amount of employment growth thus far during the current decade (e.g.,
construction and extraction occupations in the construction industry) makes
application of the BLS occupation-industry methodology problematic.
The Employment Policy Foundation (EPF) calculates a Job Quality Index,
which is the ratio of the change in employment of above-median wage jobs to the
change in employment of below-median wage jobs using a 12-month moving
average. In contrast with the use of 154 occupation-industry categories by
Factcheck.org and BusinessWeek, EPF classifies 187 occupation-industry categories
as high wage or low wage (based on 1993 earnings data). The index has been rising
since 2002.31
As with the previously discussed approaches, one shortcoming of all of these
studies is that no distinction is made among jobs within the above median wage and
below median wage groups. For example, the studies do not make any distinction
between employment growth concentrated in jobs that paid far above the median
wage and growth concentrated in jobs that paid only one dollar more than the median
wage. 32


30 Peter Coy, “Another Look at Those Job Numbers,” BusinessWeek, July 26, 2004, p. 35.
31 The Employment Policy Foundation releases the Job Quality Index on a monthly basis.
For the most recent release, see Employment Policy Foundation, “The Myth of Stagnant
Wages,” Employment Trends, July 23, 2004.
32 EPI made this argument in a rebuttal to the Factcheck.org study. EPI also criticized the
Factcheck.org study for using overly broad industry categories. Using the same data as
Factcheck.org and the same methodology as in EPI’s industry study reviewed above, it
found that between June 2003 and June 2004, contracting categories had a median weekly
wage $44 or 7% higher than expanding categories. Between June 2001 and June 2004,
contracting categories had a median weekly wage $59 or 10% higher than expanding
categories. Based on these findings, EPI concluded that expanding categories are of lower
quality than contracting categories. Elise Gould, et al., “Assessing Job Quality,” EPI Issue
Brief #200, July 28, 2004.

What Can Be Said About New Job Quality?
There are many ways to conduct a study on new job quality. The study can be
based on industry data, occupation data, or a combination of the two. A study can
compare the wages of expanding sectors to declining sectors or compare the
employment change in high wage sectors to the change in low wage sectors. A sector
can be considered to be rising or falling in absolute terms, or as a share of overall
employment. Unsurprisingly, this means that the same data can yield very different
results, as is the case with the studies discussed in this report.
When aggregating the 154 occupation-industry categories, one can lose
perspective of which categories are driving the results. For example, when results
are divided into an above median wage category and a below median wage category,
categories very close to the median wage are treated the same as categories very far
from the median. This suggests that the results would be extremely sensitive to
changes in employment in jobs close to the median wage — which BLS reports is the
reason that it has not completed an occupation-industry study utilizing recent data.
This is a drawback to nearly all the studies reviewed in this report (the EPI study is
a notable exception).
Of the 80 categories with increasing employment from June 2003-June 2004
(the same time periods used in the Factcheck.org and EPI studies), 51% of the
increase is accounted for by the 10 categories with the largest increases. Besides
making the results more manageable, focusing on the top 10 categories has the
advantage of eliminating changes in smaller categories whose results may not be
statistically significant and may be overly sensitive to the time period selected. As
seen in Table 9, construction and extraction occupations in the construction industry
(an increase of 464,000) and professional occupations in the education and health
services industry (an increase of 447,000) were by far the biggest job gainers,
accounting for 23% of the total increase between them. The former paid a median
wage very close to the overall median, and the latter paid a median wage much higher
than the overall median. The professional occupations in the education and health
services category illustrates that even when divided into 154 categories, some
categories still contain such a broad array of jobs that it is difficult to be certain
exactly what types of jobs were actually created. Of the remaining eight categories,
three were far below the median wage, two were far above the median wage, and
three were near the median wage. Six of the 10 largest increases occurred in the
education and health services industry, which has far more employment overall than
any other industry group, and the professional and business services industry group,
which contains industries paying a wide range of wages.



Table 9. Largest Increases in Employment by Occupation-
Industry Category, June 2003-June 2004
Increase inMedian
Occupation/IndustryEmploymentWeekly Wage
(thousands)2003 ($)
Construction and Extraction/Construction464553
Professional/Education and Health Services447691
Management/Financial Activities208876
Transportation/Transportation and Utilities172597
Service/Professional and Business Services167325
Sales/Professional and Business Services137506
Service/Education and Health Services127326
Office and Administrative/ Education and123453
Health Services
Installation and Repair/Professional and111632
Business Services
T r ansportation/Manufacturing 109 494
Source: Created by CRS from BLS data from the CPS.
When a similar exercise is completed for the occupation-industry categories
with the largest decreases in employment in the last year, the top 11 categories (two
categories are tied for tenth) accounted for 52% of the overall decline. (See Table
10.) The declining categories were spread more evenly across industries and
occupations than the increasing categories; the second and third largest categories
were within the manufacturing industry. Of the declining categories, three paid far
below the median wage (all of which were among the five largest declines), two paid
below but near the median wage, and six paid above the median wage (two of which
paid about double the median wage).



Table 10. Largest Decreases in Employment by
Occupation/Industry Category, June 2003-June 2004
Decrease inMedianWeekly
Occupation/IndustryEmploymentWage 2003
(thousands)($)
Office and Administrative/Retail Trade-228369
Manage me nt/Manufacturing -181 1125
Production/Manufacturing -162 509
Production/Education and Health Services-115354
Protective/Professional and Business Services-115394
Professional/Professional and Business Services-115938
Installation and Repair/Transportation and Utilities-111789
Management/Information -94 1101
Office and Administrative/Information-86497
Professional/Other services-84611
Management/Other services-84827
Source: Created by CRS from BLS data from the CPS.
Uses and Limitations of the Data
Job quality is a concept that embodies many factors, including wages, benefits,
work environment, physical safety, job security, and so on. Earnings are the only one
of these factors for which comprehensive data exist, and so they are the usual
indicator of quality. There are no data available that directly answer the question of
what wages are paid to the newly created jobs in the economy. The closest that
researchers can get to answering that question is to find out what industry or
occupation the new jobs are located in and what the average/median earnings are for
that industry or occupation. Industry and occupation categories are too broad to give
very meaningful information on job quality, however. Combining them gives more
precise information, but at the cost of statistical robustness for some of the smaller
categories.
Different studies have reached starkly opposite conclusions. Some of the
differences can be attributed to the fact that the studies covered different time
periods; some are due to differences in methodology. Looking at the occupation-
industry categories that have contributed the largest portion of the increase or
decrease in employment suggests that most of the studies have arguably drawn overly
simplified conclusions on the question of new job quality. Both the large expanding
and contracting categories fall throughout the pay spectrum: based on the data in
Tables 9 and 10, there does not appear to be a strong correlation between earnings
levels and employment gains or losses.



While these data give us some indication of the state of one segment of the labor
market, they tell us little about the overall labor market. That is, information on the
1.4 million jobs created in the past year tells us little about the overall 139 million
jobs or the 8.2 million unemployed workers in the U.S. economy. The wages of 1.4
million new jobs tell us little about how much overall earnings are rising (real wage
growth has been low since the last expansion), and they tell us little about other
important factors such as fringe benefits and working conditions.33 They also give
no indication if the distribution of earnings has become more or less equal.34
Changes in the earnings of workers who remain in the same job or who move from
one already existing job to another are likely to have a much larger effect on the
overall labor market than new jobs.
Wage data on net new jobs also do not shed light on the subsequent labor
market experience of people who lost jobs. A periodic supplement to the CPS, the
Displaced Worker Survey, asks long-tenured workers detailed questions about their
post-displacement history. Workers with at least three years of job tenure are
considered displaced if they lost a job because their plant or company closed or
moved, there was insufficient work, or their position or shift was abolished. In the
most recent survey, which covers 2001-2003, 5.3 million workers were displaced and
65% of the displaced workers were reemployed at the time of the survey. Of those
who lost full-time jobs and subsequently found new jobs, 10% were working part
time, 24% were working full time at wages that were at least 20% lower than their
previous job, and 30% were reemployed at higher wages.35 The data utilized to
assess the quality of new jobs do not address whether displaced workers were able
to obtain new jobs that are high or low paying relative to their former jobs.
Expanding and contracting occupation-industry groups might be of interest if
they accurately predicted which occupation-industry groups would expand or contract
over the entire expansion, but this may not be the case. For example, the rapid
increase in construction employment may not continue once interest rates rise. This
example suggests that, in the short run, changes in product demand are an important
determinant of where job creation occurs — and these largely uncontrollable changes
can affect workers at any wage level. It may also be that lower wage jobs are the first
to be created in an expansion, and higher wage jobs come later — as was the case in
past expansions.36 For example, employment growth in temporary help agencies
tends to lead overall employment. For the labor market as a whole, the slow rate of
employment and wage growth are more noteworthy to date in this economic
expansion than the quality of jobs being created


33 David R. Francis, “The Two Economies: The Economy is Growing and Jobs are Being
Created but Wages Don’t Keep Up,” The Christian Science Monitor, Aug. 4, 2004.
34 For more information, see CRS Report RL31616, The Distribution of Earnings of Wage
and Salary Workers in the United States, 1994-2003, by Gerald Mayer.
35 Bureau of Labor Statistics, Worker Displacement, 2001-2003, news release USDL04-

1381, July 30, 2004.


36 Ilg and Haugen, Earnings and Employment Trends in the 1990s and Aaronson and
Christopher, Employment Growth in Higher-Paying Sectors.