13q1tuzemen willis 1

7/30/2019 13q1Tuzemen Willis 1

1/28

The Vanishing Middle:Job Polarization and WorkersResponse to the Decline inMiddle-Skill JobsBy Didem Tzemen and Jonathan Willis

Didem Tzemen is an economist at the Federal Reserve Bank of Kansas City. JonathanL. Willis is an assistant vice president and economist at the bank. Shujaat Khan, aresearch associate, helped in the preparation of the article. This article is on the bankswebsite atwww.KansasCityFed.org.

5

Over the past three decades, the share o middle-skill jobs in the

United States has allen sharply. Middle-skill jobs are those inwhich workers primarily perorm routine tasks that are proce-

dural and repetitive. The decline in the employment share o middle-

skill jobs has been associated with a number o sweeping changes aect-ing the economy, including advancement o technology, outsourcing o

jobs overseas, and contractions that have occurred in manuacturing.

As the share o middle-skill jobs has shrunk, the share o high-skill jobshas grown, and that trend has drawn considerable attention. Less wellknown is the act that the share o low-skill jobs has also risen. This

employment phenomenon where job opportunities have shited awayrom middle-skill jobs toward high- and low-skill jobs is called jobpolarization.

The impact o job polarization is not well understood. Researchershave oered various theories to explain job polarization and its rela-

tion to the business cycle, but there have been only limited studies onthe impact o job polarization across industries and across various seg-ments o the population. The knowledge gap to be resolved is how jobpolarization has aected the structure o the labor market. Understand-


2/28

6 FEDERAL RESERVE BANK OF KANSAS CITY

ing this changing structure o the labor market can help policy makers

prescribe policies that will best promote sustainable economic growth.

This article examines three decades o data rom the Current Popu-lation Survey to characterize structural changes in the labor market due

to job polarization. One common assumption is that job polarizationhas been driven mainly by contraction in a ew sectors, such as manu-acturing, where a large share o jobs are in the middle-skill category.

However, the analysis shows that job polarization has stemmed lessrom a shit away rom any one sector than rom shits in the skill-composition o jobs within each sector. Related to the business cycle,

job polarization has occurred consistently over the last three decades,but the pace o job polarization has accelerated during recessions.The article also nds that workers response to job polarization has

diered depending on their gender, education, and age. Women haveobtained more education and moved disproportionately into high-skill

jobs, while men have shited in roughly equal numbers to high- and

low-skill jobs. Workers age 55 and older have shited strongly towardhigh-skill jobs, as workers in high-skill occupations have delayed retire-ment. For the youngest segment o the labor market, workers ages 16to 24 have shited toward low-skill jobs as a growing segment o this

population have delayed entry into the labor market while remainingin school.

Section I describes the course o job polarization over the past three

decades in the United States and reviews various theories oered to ex-plain the pattern. Section II shows how job polarization occurred acrossdierent sectors o the economy, ocusing on the demand side o the

labor market. Section III ocuses on the supply side o the labor market,

nding dierences across workers based on gender, education, and age.Section IV explores the relationship between job polarization and the

business cycle.

I. JOB POLARIZATION IN THE U.S. LABOR MARKET

The term job polarization has been used in studies about bothBritish employment (Goos and Manning) and U.S. employment (Au-tor and others, 2006) to describe the shit o the workorce toward

low- and high-skill occupations. A variety o explanations have beenoered or the pattern, including the impact o technology, the rise in


3/28

ECONOMIC REVIEW FIRST QUARTER 2013 7

international trade, and the shrinking and weakening o labor unions.

These actors have aected the demand side o the labor market,

infuencing the types o jobs that are in demand at a given time. How-ever, job polarization has also been refected in the response o workers,

corresponding to the supply side o the labor market. Both supply- anddemand-side actors are considered in this study.

The shit in the composition o jobs in the U.S. labor market is

documented using micro-level data rom the Current Population Sur-vey (CPS), commonly reerred to as the household survey. The U.S.Census Bureau collects survey data at a monthly requency rom ap-

proximately 60,000 households. The data contain detailed demograph-ic inormation on workers, as well as inormation on employment byindustry and occupation. The analysis in this article is based on annual

data rom 1983 to 2012, except in the nal section where quarterly dataare used or the business-cycle analysis.1

Using the CPS micro-level data, this section rst measures the ex-

tent o job polarization in the United States over the past thirty yearsand reviews theories that seek to explain the trend. The section thenexamines the evidence rom wage patterns to explore whether laborsupply has shited to meet new patterns o labor demand.

Explanations for job polarization

Changes in technology and the global economy together haveprovided the impetus or job polarization. With the adoption ocomputers and associated technologies, businesses have undamentally

changed the way they operate and the types o workers they require.Simultaneously, international trade has expanded rapidly, leading to newopportunities or some companies and increased competition or oth-

ers. U.S. industries that were not competitive in the global marketplacehave contracted. The decline in labor unions may have also contributedto job polarization through a weakening o the bargaining power o

workers predominantly in middle-skill jobs. Job polarization has thusbeen the product o a wide variety o orces that have shited employ-ment toward more productive occupations in a perpetually changingeconomic landscape.

Beginning in the 1980s, technological advancements and the wide-spread adoption o computers have led to a rise in the relative demand


4/28


or workers in high-skill occupations. Workers suitable or these posi-tions are typically highly educated and can perorm tasks requiring ana-

lytical ability, problem solving, and creativity. They work at managerial,proessional, and technical occupations, such as engineering, nance,management, and medicine.2 Economists have reerred to this demandshit as skill-biased technical change (SBTC) (Goldin and Katz and

Krusell and others).The SBTC hypothesis is successul in explaining the rise in the

employment share o workers in high-skill jobs over the past three de-

cades. This hypothesis suggests that new technologies complement the

skill sets o highly skilled workers, increasing their relative productiv-ity and leading to the creation o more high-skill jobs.3 More precisely,

the employment share o workers in high-skill occupations increased by11 percentage points rom 26 percent in 1983 to 37 percent in 2012(Chart 1).

However, the SBTC hypothesis ails to explain the increase in theshare o low-skill occupations. Workers in these jobs typically have noormal education beyond high school. They work in occupations that

are physically demanding and cannot be automated. Many o these oc-cupations are service oriented, such as ood preparation, cleaning, andsecurity and protective services. The SBTC hypothesis predicts that

the share o low-skill jobs should have allen. Instead, the employmentshare o low-skill occupations rose rom 15 percent in 1983 to 18 per-cent in 2012.

While the employment shares have increased or low- and high-skill occupations, the employment share has declined or middle-skilloccupations. These middle-skill occupations include sales, oce and

administrative support, production, construction, extraction, instal-lation, maintenance and repair, transportation, and material moving.From 1983 to 2012, the employment share o middle-skill occupationsdropped by 14 percentage points (rom 59 percent to 45 percent).

Technological advancements help explain why the share o workersemployed in middle-skill occupations has allen so sharply. Workers inmiddle-skill occupations typically perorm routine tasks that are pro-

cedural and rule-based. Thereore, these occupations are classied as

routine occupations. The tasks perormed in many o these occupa-tions have become automated by computers and machines (Autor and


5/28


others, 1998, 2003; Acemoglu and Autor). In contrast, tasks perormed

in high- and low-skill occupations cannot be automated, making themnon-routine occupations. Thus, the technical change that boosted

the demand or high-skill jobs also contributed to the all in demandor middle-skill jobs, as computers and machines became cost-eectivesubstitutes or these workers.

International trade and the weakening o unions have also con-

tributed to the decline in middle-skill occupations. Many jobs in thiscategory, particularly those in manuacturing, have been oshored tocountries where workers can perorm similar tasks or lower wages

(Autor, Goos and others, and Oldenski). In addition, some rms have

been contracting out portions o their businesses to workers in oreigncountries, a phenomenon known as outsourcing. De-unionizaton has

also potentially contributed to the decline in the employment share omiddle-skill jobs (Autor).

Job polarization and wages

Evidence rom U.S. wage patterns suggests that the sup-ply o workers, in stride with labor demand, has shited away rom

middle-skill occupations. According to the theory o supply anddemand, a decrease in the relative demand or workers in middle-skilloccupations results in a decline in the relative wage or those workers.

Note: Data are restricted to workers ages 16 to 64 who are not sel-employed and are not employed in military oragricultural occupations.Source: Current Population Survey and authors calculations.

Chart 1

EMPLOYMENT SHARES BY SKILL LEVEL

Low-SkillOccupations 15%

Middle-SkillOccupations59%

High-SkillOccupations

37%

Low-SkillOccupations

18%

Middle-Skill

Occupations45%

1983 2012

High-SkillOccupations26%


6/28


Similarly, an increase in the relative demand or workers in low- and

high-skill occupations leads to higher relative wages or these workers.

Wage dynamics that do not ollow these patterns imply shits in therelative supply o workers. An analysis o wage data nds that relative

wages across skill level categories have changed very little over the past

30 years, suggesting that labor supply has moved in the same directionand roughly in line with changes in labor demand.

Based on the earnings inormation provided by the CPS, wag-

es or occupations in one skill level can be compared with wages oroccupations in other skill levels. For each skill category, a measure o the

relative wage is constructed by calculating the ratio o the median wagein that category to the median wage or all workers.4 As expected, relativewages are closely associated with skill levels (Chart 2). The relative wageor workers in high-skill occupations is the highest. The relative wage or

workers in middle-skill occupations is near 1, indicating that the medianwage or this skill category is very close to the overall median wage. Rela-tive wages or workers in low-skill occupations are the lowest.

Wage patterns suggest that labor supply has shited in response tojob polarization. For workers in middle-skill occupations, the relativewage has remained very stable despite the 14 percentage-point decrease


Chart 2

MEDIAN WAGES FOR WORKERS BY SKILL LEVEL

0.4

0.6

0.8

1

1.2

1.4

1.6

0.4

0.6

0.8

1

1.2

1.4

1.6

1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011

Median Wages Relative to the Median Wage Across All Workers, Ratio

Low-Skill Occupations

Middle-Skill Occupations

High-Skill Occupations


7/28


in the employment share o these occupations. Without any labor supply

response to this decrease, economic theory would predict a signicant

decline in relative wages. The stability o relative wages or workers inmiddle-skill occupations suggests that the decline in labor demand has

been approximately matched by a decline in labor supply or these oc-cupations.

For workers in low- and high-skill occupations, relative wage patterns

are also inormative regarding labor supply movements. The employ-ment share o workers in low-skill occupations increased by 3 percentagepoints, while the employment share o workers in high-skill occupations

increased by 11 percentage points over the sample period. The stablepatterns or relative wages or both low- and high-skill occupations sug-gest that the increases in labor demand or these occupations have been

accompanied by similar increases in labor supply.

II. CHANGES IN LABOR DEMAND:

JOB POLARIZATION ACROSS SECTORS

The shiting o employment away rom middle-skill occupations

in the aggregate can occur in two ways. First, employment may shitbetweensectors o the economy as some sectors experience increases indemand or their products and respond by hiring workers, while other

sectors contract in the ace o weakening demand or their products.Worker reallocation contributes to job polarization i the expandingsectors have a larger share o low- or high-skill jobs and the contractingsectors have a larger share o middle-skill jobs.

Second, actors underlying job polarization may also contribute toa shit o employment withina sector as middle-skill jobs are eliminated

and the shares o low- and high-skill jobs rise. Technological advance-ments tend to be the primary drivers o this aspect o job polarization,as middle-skill jobs are made obsolete. Examples o within-sector shitsinclude the replacement o secretaries with computers and the replace-

ment o middle-skill actory jobs with advanced machinery and robots.Distinguishing between these two types o shits in employment

patterns is important or the ecacy o labor market policy. I job

polarization is driven mainly by the contraction o sectors with larger

shares o middle-skill jobs, then labor market policy needs to ocus onhelping displaced workers in shrinking sectors make the transition to

other sectors where employment opportunities are prolierating. In


8/28


contrast, i job polarization is driven mainly by within-sector shits in

employment, then labor market policy should help provide incentives

and opportunities or workers in middle-skill jobs to obtain the neces-sary skills to become qualied or high-skill jobs.

The initial analysis reveals two important contributors to job po-larization. First, the employment shares o sectors that have higher pro-portions o middle-skill jobs have declined, and the employment shares

o sectors with lower proportions o middle-skill jobs have increased.Second, all sectors experienced decreases in the shares o middle-skill

jobs rom 1983 to 2012.

A decomposition is used to identiy whether employment shits be-tweenor within sectors are most responsible or job polarization. Thedecomposition indicates that the decline in middle-skill jobs within sec-

tors has contributed more to job polarization than shits in employmentbetween sectors.

Employment shifts across sectors

Job polarization has taken place during a period o large shits in em-

ployment across industries. Analyzing these shits in overall employmentand the corresponding changes in skill-level compositions o jobs withineach industry reveals that overall employment has shited away rom in-dustries with larger shares o middle-skill jobs and toward industries with

larger shares o low- and high-skill jobs. Also, the analysis shows that theshares o middle-skill jobs across all industries have declined.

For purposes o analysis, the economy is divided into our sectors.

The rst two sectors are construction and manuacturing, where middle-skill occupations comprise more than 70 percent o employment. Thethird sector groups together the education and health services industries,

where high-skill occupations account or the largest share o employ-ment. The nal sector, labeled other, encompasses all other industries,a group in which middle-skill occupations account or 50 percent o

employment. This sector includes mining; wholesale and retail trade;transportation and utilities; communication; inormation; nancial ac-tivities; other services; public administration; agriculture; and orestry,shing, and hunting.

From 1983 to 2012, the employment shares o most sectors haveshited markedly. The largest shit has been the decline in manuactur-


9/28


ing, where the employment share ell rom 22 percent in 1983 to 11percent in 2012 (Chart 3). For the education and health services sec-tor, the employment share increased rom 17 to 24 percent over the

period. The employment share or the other sector rose rom 55 to60 percent, while the employment share or the construction sector wasbasically unchanged.

The share o workers in middle-skill occupations declined acrossall o these sectors. The declining employment share is measured bytracking the change in the employment shares o occupations within

each sector in each o the three skill levels rom 1983 to 2012. The sec-tor with the largest decline in middle-skill jobs was the other sector,

which experienced an 11 percentage-point decrease in the employmentshare o workers in middle-skill jobs (Chart 4). The remaining threesectors experienced declines in the employment share o workers inmiddle-skill occupations ranging rom 6 to 9 percentage points.

For these three sectors, the decline in the employment share oworkers in middle-skill occupations was primarily oset by increasesin the share o workers in high-skill occupations. The construction,manuacturing, and education and health services sectors experienced

increases in the employment share o workers in high-skill occupationso the same magnitude as the decline o those in middle-skill occu-pations, while the share o workers in low-skill occupations remained

Chart 3

EMPLOYMENT SHARES BY SECTOR

1983 2012

Other55%

Construction 6%

Manufacturing22%

Educationand Health17%

Other60%

Construction 5%

Manufacturing11%

Educationand Health24%



10/28


unchanged. For the other sector, the share o workers in high-skill oc-

cupations increased by 8 percentage points, while the share o workersin low-skill occupations increased modestly.

Decomposing employment shifts within and between sectors

The primary actor contributing to job polarization can be deter-mined by identiying the underlying source o the observed shits in

employment across sectors or workers in each skill level. Specically,the employment shits are decomposed into two components: skill-based employment shits within sectors and skill-based employment

shits betweensectors.The rst component is the portion o job polarization associated

with changes in the skill-based composition o jobs withineach sector.

This component captures shits in the composition o jobs, holdingxed the size o each sector.5 It shows how much o the observed change

in the employment share o workers in middle-skill jobs in a sector is

Chart 4

WITHIN-SECTOR EMPLOYMENT SHARES BY SKILL LEVEL

Note: Employment shares are computed separately or each sector. Data are restricted to workers ages 16 to 64 whoare not sel-employed and are not employed in military or agricultural occupations.Source: Current Population Survey and authors calculations.

10

20

30

40

50

60

70

80

90

100

10

20

30

40

50

60

70

80

90

100

1983 2012 1983 2012 1983 2012 1983 2012

Construction Manufacturing Education and Health Other

Percentage

Low-Skill Occupations Middle-Skill Occupations High-Skill Occupations


11/28


due to a reallocation o workers to or rom low- and high-skill jobs

within that sector.The second component measures the portion o job polarization

associated with shits in employment betweensectors. The measure is

computed based on the observed change in each sectors share o to-tal employment, holding constant the skill-based composition o jobs

within each sector. This component conveys how much o job polar-

ization is due to a reallocation o workers rom contracting sectors toexpanding sectors.

As the main theme o job polarization is the vanishing o middle-skill jobs, the analysis begins with the decomposition o the decline inthe employment share o workers in middle-skill occupations. The laterpart o the analysis ocuses on the decomposition o the corresponding

increases in low- and high-skill jobs.Based on this decomposition, the primary contributor to the

decline in middle-skill jobs was the change in the skill-composition o

jobs within sectors. The overall decline in the employment share o work-

ers in middle-skill occupations was 14 percentage points rom 1983 to2012 (Table 1). Within-sector changes in the skill composition o jobs

Table 1

CHANGES IN EMPLOYMENT SHARES BY SKILL LEVEL

AND SECTOR BETWEEN 1983 AND 2012

Note: The table reports the percentage point change in the employment share or each skill group, where theemployment share is measured as the ratio o the skill groups employment to total employment. Data are restrictedto workers ages 16 to 64 who are not sel-employed and are not employed in military or agricultural occupations.For each sector, the employment share in 2012 is shown in parentheses.Source: Current Population Survey and authors calculations.

Total Construction Manuacturing Education& Health

Other

Skill level (5%) (11%) (24%) (60%)

Low-skill occupations 3.7 0.0 -0.2 1.4 2.6

Change due to shits between sectors 2.1 0.0 -0.2 1.5 0.9

Change due to shits within sector 1.6 0.0 0.0 -0.1 1.7

Middle-skill occupations -14.2 -0.4 -9.7 -0.4 -3.8

Change due to shits between sectors -4.5 0.0 -8.7 1.5 2.8

Change due to shits within sector -9.8 -0.3 -1.0 -1.9 -6.6High-skill occupations 10.5 0.3 -1.3 5.6 5.9

Change due to shits between sectors 2.3 0.0 -2.3 3.6 1.0

Change due to shits within sector 8.2 0.3 1.0 2.0 4.9


12/28


accounted or two-thirds, or 10 percentage points, o this decline. Theremaining one-third o the decline was due to shits in employment

between sectors, moving rom sectors with higher shares o workersin middle-skill jobs to sectors with lower shares o workers in middle-skill jobs.

Within each sector, job polarization occurred through a decline in

the share o workers in middle-skill jobs. The largest contributor waswithin the other sector, where the overall share o workers in mid-dle-skill occupations ell 7 percentage points. The other three sectors

contributed less. The combined impact o compositional shits within

these sectors accounted or two-thirds o the reduction in the employ-ment share o workers in middle-skill occupations.

Shits between sectors were a smaller portion o job polarization.The sharp decline in manuacturing employment led to a decline o 9percentage points in the overall share o workers in middle-skill jobs

rom 1983 to 2012. But this sharp decline in middle-skill jobs waspartially dampened as employment shited to middle-skill jobs in theother three sectors. The increase in the employment share o the oth-

er sector, controlling or compositional shits o workers in this sector,added 3 percentage points to the overall share o workers in middle-skill jobs. The contribution rom expanding employment in the educa-

tion and health services sector was positive, but smaller, as this sectorhad a smaller share o workers in middle-skill occupations. On net, theshit o workers over the past three decades away rom manuacturing

and toward the other three sectors accounted or about one-third o thereduction in the overall share o workers in middle-skill jobs.

Similar to the decline in middle-skill occupations, the increase in

the employment share o workers in high-skill occupations was pri-marily due to within-sector shits. All sectors experienced within-sectorincreases in the share o workers in high-skill jobs, with the largestincrease occurring in the other sector. Shits in employment between

sectors accounted or a smaller increase. As employment shited awayrom manuacturing and toward the education and health services sec-tor and the other sector, the share o workers in high-skill jobs in-

creased. This was because the latter sectors had higher shares o workers

in high-skill jobs than manuacturing.The share o workers in low-skill occupations rose moderately due

to similar contributions rom within- and between-sector shits in the


13/28


employment shares. The shit in employment away rom manuactur-

ing increased the share o workers in low-skill jobs because the shit

was toward sectors with larger concentrations o workers in low-skilljobs. The shits within the other sector also increased the employ-

ment share o workers in low-skill occupations.In summary, job polarization is primarily attributable to within-

sector shits in employment. In all sectors the employment share o

middle-skill jobs declined, and the share o high-skill jobs increased.This within-sector shit in employment accounted or two-thirds othe reduction in the share o middle-skill jobs, while the remaining

one-third was due to shits in employment between sectors. The im-pact o the sharp decline in manuacturing employment on middle-skill jobs was partially oset by an increase in the share o workers in

middle-skill jobs in the remaining sectors.

III. CHANGES IN LABOR SUPPLY: WORKERS

RESPONSE TO JOB POLARIZATION

As job polarization has shited employment opportunities, workershave responded. Faced with increasing demand or workers in high-skill

occupations and decreasing demand or workers in middle-skill occupa-tions, individuals have a strong incentive to attain higher levels o edu-cation. But not all individuals respond in the same way. Responses may

depend on characteristics such as age, gender, and level o education.Understanding how workers react to job polarization is important ordesigning and implementing eective labor market policy.

Using the CPS data, a detailed analysis o the labor supply responseto job polarization reveals dierences across workers. Women have re-

sponded to job polarization by attaining higher levels o education andshiting predominantly to high-skill occupations. In constrast, menhave responded to the decline in middle-skill jobs with a more modestincrease in education and roughly equal shits toward low- and high-

skill jobs. Workers age 55 and over have shited exclusively to high-skill occupations while prime-age workers, ages 25 to 54, have shitedto both low- and high-skill occupations. Among young people in the

labor orce, workers ages 16 to 24 have shited primarily to low-skill

occupations. This shit was likely related to the increasing number o


14/28


15/28


tirely refected in a 16 percentage-point increase in the share o employed

women in high-skill jobs. In contrast, during this time period the share oemployed women in low-skill jobs increased only modestly.

Unlike women, male employment shares have shited equally to-

ward low- and high-skill jobs. In 1983, 62 percent o employed menwere in middle-skill jobs, while the shares o employed men in low- andhigh-skill jobs were 27 percent and 11 percent, respectively. By 2012,

the share o employed men in middle-skill jobs ell by 10 percentagepoints, while employment shares o low- and high-skill jobs each roseby 5 percentage points.

To better assess gender dierences, it is useul to consider employ-ment shits among major occupations within each skill group. For women

most o the employment gains were in managerial and proessional occu-pations, which are classied as high-skill jobs (Table 2). For men the em-ployment gains in these occupations were modest. Both men and womensaw declines in middle-skill jobs, but or dierent reasons. Middle-skill

jobs are in routine occupations such as sales, oce and administrative,production, construction, installation, maintenance, and transportation.The main losses in the share o middle-skill employment or women were

in oce and administrative occupations, likely related to the replacement

o many secretarial and clerical jobs with desktop computing. Major em-ployment losses or men were in production occupations, likely related to

the decline in manuacturing employment.

Table 2

EMPLOYMENT SHARES OF OCCUPATIONS BY GENDER

Note: Employment shares are computed separately or each gender and expressed in percentages. Data are restrictedto workers ages 16 to 64 who are not sel-employed and are not employed in military or agricultural occupations.Source: Current Population Survey and authors calculations.

Men Women

Occupations 1983 2012 Change 1983 2012 Change

Management, Business, and Financial 11.9 14.4 2.5 7.8 14.1 6.4

Proessional and Related 15.1 18.1 3.0 18.0 27.4 9.5

Service 11.2 16.0 4.8 18.7 20.9 2.3

Sales and Related 9.5 10.0 0.5 11.8 11.0 -0.8

Oce and Administrative Support 6.8 7.2 0.3 31.2 20.2 -11.0

Construction, Extraction, Installation,Maintenance, Repair, and Production

37.8 24.1 -13.6 11.9 4.1 -7.7

Transportation and Material Moving 7.7 10.2 2.5 0.8 2.1 1.4


16/28


For low-skill occupations, the employment gains were skewed

toward men. These jobs were primarily in service occupations. For

both men and women, services employment increased, but gains werestronger or men.

To summarize, job polarization has aected both men and wom-en, but they have responded dierently. For both sexes, the share oemployment in middle-skill occupations ell dramatically. This decreaseled to equal increases in the shares o male employment in low- and

high-skill occupations. However, almost all the decrease in the emaleemployment share in middle-skill occupations was transormed into a

large increase in the employment share o women in high-skill occupa-tions, with almost no change in their share in low-skill occupations.The ndings o several recent studies may help explain the dier-

ent responses o men and women to job polarization. As the share o

middle-skill jobs declined, workers were conronted with a choice o tak-ing low-skill jobs or pursuing the skills necessary or high-skill jobs. Heretwo actors came into play. First, low-skill jobs paid lower wages (Chart

2). Second, employment decisions o men were less sensitive to wagesthan those o women (Kimmel and Kniesner). Rather than accepting alower wage, women were more likely to pursue education and obtain em-

ployment in high-skill occupations. Moreover, women, and particularlycollege-educated women, were likely to be more responsive to increaseddemand or high-skill occupations than men (Black and Juhn).

Job polarization and educational attainment

The shit in employment toward high-skill jobs has been associated

with an increase in the share o workers obtaining higher levels o edu-cation. In 1983, 15 percent o the working population had less than ahigh school degree (Chart 6). This share declined to 9 percent in 2012.

Over the same period, the share o workers with a high school diplomaell rom 40 percent to 27 percent. In contrast, or workers with somecollege education or an associates degree, the share increased rom 22

percent to 30 percent, and or those with a bachelors degree or more,the share rose rom 23 percent to 34 percent.

Job polarization has disproportionally aected workers with less

than a bachelors degree. In 1983, the majority o workers in theseeducational categoriesless than high school, high school degree, andsome college or associate degreewere employed in middle-skill occu-


17/28


Table 3

EMPLOYMENT SHARES OF WORKERS BY LEVELOF EDUCATION

Note: Employment shares are computed separately or each respective level o educational attainment and expressedin percentages. Data are restricted to workers between the ages 16 to 64 who are not sel-employed and are notemployed in military or agricultural occupations.Source: Current Population Survey and authors calculations.

Level o Educational Attainment Occupation Type 1983 2012 Change

Low-Skill Occupations 26.9 36.9 10.0

Less Than High School Middle-Skill Occupations 69.2 58.1 -11.1

High-Skill Occupations 3.9 5.0 1.1


High School Degree Middle-Skill Occupations 72.6 62.4 -10.3



Some College or Associate Degree Middle-Skill Occupations 60.7 50.8 -9.9



Bachelors Degree or Higher Middle-Skill Occupations 26.0 21.4 -4.6


Chart 6

EDUCATIONAL ATTAINMENT OF THE EMPLOYED


Less thanHigh School15%

High SchoolDegree40%

Some Collegeor Associate Degree22% Some College or

Associate Degree30%

Bachelors Degreeor Higher34%

Less thanHigh School9%

HighSchoolDegree27%

1983 2012

Bachelors Degreeor Higher23%


18/28


pations (Table 3). By 2012, all three groups in middle-skill occupations

experienced employment-share declines o approximately 10 percentagepoints. Employment gains or these groups were primarily in low-skilloccupations as these individuals did not have the education and training

required or high-skill occupations.Workers with a bachelors degree or higher have been less aected

by job polarization. In 1983, only 26 percent o these workers were in

low- and middle-skill occupations. From 1983 to 2012, the share oworkers with a bachelors degree or higher in middle-skill jobs declined

modestly by 5 percentage points. This decline was accompanied by asmall increase in the employment share o these workers in low- andhigh-skill occupations.

While levels o educational attainment increased or both male and

emale workers, the change was more pronounced or women. From1983 to 2012, the share o employed women with a high school degreeell dramatically, rom 43 percent to 25 percent (Chart 7). Over the same

period, the share o emale workers with a bachelors degree or higher

rose. For men, the share o workers with a high school degree ell moremodestly, rom 38 percent to 29 percent. This decline was accompaniedby comparable increases in the share o male workers with some college

or an associates degree and the share with a bachelors degree or higher.

Chart 7

EDUCATIONAL ATTAINMENT OF THE EMPLOYED BY GENDER

5

10

15

20

25

30

35

40

45

5

10

15

20

25

30

35

40

45Men Women

Percentage

1983 2012

Less thanHigh School

High SchoolDegree

Some Collegeor Associate Degree

Bachelors Degreeor Higher

Less thanHigh School

High SchoolDegree

Some Collegeor Associate Degree

Bachelors Degreeor Higher

Note: The chart reports levels o educational attainment separately or each respective gender. Data are restricted toworkers ages 16 to 64 who are not sel-employed and are not employed in military or agricultural occupations.Source: Current Population Survey and authors calculations.


19/28


Gender dierences in educational attainment line up closely withshits in the employment shares by skill level. The large increase rom 1983to 2012 in the share o women with a bachelors degree or higher was

accompanied by strong employment gains or women in high-skill occu-pations. For men, the smaller increase in educational attainment matchedthe smaller shit o male employment toward high-skill occupations.

Job polarization and age

Two important trends have changed the age prole o the labororce. First, the aging o the baby boom generation has led to an olderworkorce. Second, the increase in the demand or workers with higherlevels o educational attainment has led younger people to stay in school

longer and highly-educated older workers to delay retirement.These trends have contributed to a decline in the employment share

o younger workers and an increase in the employment share o older

workers. As younger individuals have spent more time in school, theemployment share o workers ages 16 to 24 declined rom 22 percent to14 percent rom 1983 to 2012 (Chart 8). Over the same timerame, the

Chart 8

AGE DISTRIBUTION OF THE EMPLOYED


Ages 16-2422%

Ages 25-5467%

Ages 55+11% Ages 16-24

14%

Ages 25-5470%

Ages 55+16%

1983 2012


20/28


21/28


IV. JOB POLARIZATION OVER THE BUSINESS CYCLE

While there is wide agreement among researchers regarding the

emergence o job polarization in the U.S. labor market, there is no con-sensus yet about the relationship between job polarization and the busi-

ness cycle. I skill-based shits in the employment shares occurred con-sistently over time, job polarization may be seen as a structural changethat is independent rom business-cycle fuctuations. I the shits have

been concentrated in periods o economic contraction or economic ex-pansion, job polarization may be seen as closely tied to the businesscycle. This distinction is important because it may lead to an improved

understanding o recent patterns in the labor market, such as the emer-gence o jobless recoveries. The ramework o analysis used in this articlecan provide new insights on this open question.

Thus ar, economic studies have presented mixed evidence regard-ing the connection between job polarization and the business cycle.One recent study ound that nearly all o the job losses occurring during

periods o economic downturn were in middle-skill occupations, andjob losses in middle-skill occupations were concentrated in downturns

and sluggish recoveries (Jaimovich and Siu). The study concluded thatjob polarization was a key driver o recent business cycles and accountedor periods o jobless recovery that have ollowed the past three reces-sions. A second study reached a dierent conclusion, nding that reces-

sions adversely aected all workers, even those in high-skill occupations(Foote and Ryan). This study showed that workers in middle-skill oc-cupations were no more likely to leave the labor orce or take low- or

high-skill jobs in recessions than they were during economic booms.

This evidence suggests that job polarization was more likely a contribu-tor rather than the cause o jobless recoveries.

This article provides new evidence on this debate by examiningwhether shits in the employment shares associated with job polariza-tion were concentrated in recessions. For improved detection o businesscycle patterns, the CPS data used in preceding sections is converted into

quarterly data using the same methodology described in Section 1. Theanalysis reveals that job polarization occurred continuously and was notconcentrated in a ew periods (Chart 10).7 The ndings indicate that

over the past three decades skill-based shits in the employment shares


22/28


occurred consistently across periods o business cycle expansion and

contraction. This evidence suggests that job polarization is best seen asa structural phenomenon rather than a phenomenon tied to the busi-

ness cycle or associated with jobless recoveries.While job polarization has been predominantly a structural phe-

nomenon, closer analysis reveals some cyclical eatures. Shits in the

employment shares associated with job polarization have dispropor-tionally taken place during periods o economic recession. O the 118quarters spanning the period rom the rst quarter o 1983 to the third

quarter o 2012, the economy was in recession in 11 quarters or 9 per-cent o the time.8 Approximately 35 percent o the observed decline inthe employment share o workers in middle-skill occupations occurredduring these recessionary quarters. So while the majority (65 percent)

o job polarization took place during periods o economic expansion,a disproportionately large share o the shit in the composition o oc-cupations occurred during the ew periods o recession.

Taking these numbers together suggests that job polarization

occurred at a aster pace during recessions. To measure the speed ojob polarization over the business cycle, the rate o change o the

Chart 10

EMPLOYMENT SHARES BY SKILL LEVEL


10

20

30

40

50

60

70

80

90

100

10

20

30

40

50

60

70

80

90

100

1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011

Percentage

Low-Skill Occupations Middle-Skill Occupations High-Skill Occupations


23/28


Table 4

CYCLICAL CHANGES IN EMPLOYMENT SHARES BY

SKILL LEVELTotal Construction Manuacturing Education &

HealthOther

Skill level (5%) (11%) (24%) (60%)

Low-skill occupations

During recessions(4-quarter rate o change)

0.6 0.0 0.0 0.3 0.3

During expansions(4-quarter rate o change)

0.1 0.0 0.0 0.0 0.1

Middle-skill occupations


-1.7 -0.6 -1.1 0.1 -0.3


-0.3 0.1 -0.2 0.0 -0.1

High-skill occupations


1.2 0.0 0.0 0.8 0.3


0.2 0.0 -0.1 0.1 0.2

Note: This table reports the percentage point change in the employment share or each skill group, where theemployment share is measured as the ratio o the skill groups employment to total employment. Data rom 1983

to 2012 are used to construct changes in employment shares. Recessionary periods are based on NBER businesscycle dates. Data are restricted to workers ages 16 to 64 who are not sel-employed and are not employed inmilitary or agricultural occupations. For each sector, the employment share in 2012 is shown in parentheses.Source: Current Population Survey and authors calculations.

employment shares over 4-quarter periods was computed separately orrecessions and expansions. During recessions, the employment shareo workers in middle-skill occupations declined at a 4-quarter rate o

1.7 percentage points. In contrast, during expansions the employmentshare o workers in middle-skill occupations declined at a slower rate

o 0.3 percentage points. The six-old acceleration o job polarizationduring recessions relative to expansions also corresponded to the ac-celerated increases in the employment shares o workers in low- andhigh-skill occupations.

The rapid pace o job polarization during recessions can be traced toshits in the manuacturing and construction sectors. During recessions,the bulk o the decline in the share o middle-skill occupations occurred

in the manuacturing sector. The decline stemming rom changes in the

construction sector was about hal as large (Table 4). The other sector,which primarily consists o service industries excluding education and


24/28


health services, contributed modestly to job polarization. As these threesectors shited away rom middle-skill occupations, the biggest employ-

ment share increase during recessions was or workers in high-skill oc-cupations in the education and health services sector. Smaller increasesoccurred during recessions or the employment shares o workers inlow-skill occupations in the education and health services sector and or

workers in low- and high-skill occupations in the other sector.During expansions skill-based shits occurred ar more gradually.

The shares o workers in middle-skill jobs in the manuacturing and the

other sectors declined during expansions, while the shares o workers

in high-skill jobs in the education and health services sectors graduallyincreased.

Skill-based shits accelerated during recessions in the manuactur-ing, construction, and education and health services sectors. For themanuacturing sector and the education and health services sector,

skill-based shits were in the same direction during recessions and ex-pansions, but the pace o job polarization was six- to eight-times asterduring recessions. Construction contributed to job polarization during

recessions, but in expansions this sector experienced an increase in theoverall share o workers in middle-skill jobs. The other sector, in con-trast, experienced a steady shit across the business cycle in the overall

employment share away rom workers in middle-skill jobs, and towardworkers in both low- and high-skill jobs.

In summary, job polarization was evident in skill-based shits in the

employment shares in both recessions and expansions, suggesting that itis primarily a structural phenomenon. The continuous nature o shitsin the employment shares associated with job polarization suggests that

this phenomenon is neither a key driver o the business cycle nor theprimary cause o jobless recoveries. However, job polarization did dis-play some cyclicality in that the pace o job polarization acceleratedduring recessions. The construction, manuacturing, and education and

health services sectors contributed to the aster pace o job polarizationduring recessions. The other sector, which accounts or over hal ototal employment, exhibited a much more stable pace o job polariza-

tion across recessions and expansions.


25/28


V. CONCLUSION

As employment in the U.S. economy has expanded over the past

three decades, the composition o jobs has shited markedly. Job polar-ization has produced strong relative shits in the employment shares o

workers away rom middle-skill jobs and toward low- and high-skill jobs.This article shows that job polarization has been an economy-wide

phenomenon, observed in all sectors. Given the sharp decline in manu-

acturing employment in the past three decades, this sector might appearto have been the main driver o job polarization. However, empirical evi-dence reveals that job polarization has been primarily due to shits in the

skill-composition o jobs withinsectors as opposed to the shits in em-ployment betweensectors in the economy. All sectors have experienced de-clines in the within-sector share o workers in middle-skill jobs. The large

decline in middle-skill jobs associated with the contraction in manuac-turing was not the primarily driver o job polarization because this declinehas been partially oset by workers shiting to middle-skill occupations in

expanding sectors, such as the education and health services sector. Thisdistinction is important or labor market policy as it suggests that the

impact o job polarization has been widespread across the economy ratherthan concentrated in a single sector, such as manuacturing.

Job polarization has aected male and emale workers dierently.In response to the decline in the employment share o middle-skill oc-

cupations, employment o women has skewed toward high-skill occu-pations, while employment o men has shited proportionally towardlow- and high-skill occupations. The rising employment share o wom-

en in high-skill occupations has been supported by signicant increases

in educational attainment o women. Educational attainment o menhas risen more modestly, in line with the modest shit o male employ-

ment toward high-skill jobs. These shits may have important implica-tions or the skill set o the work orce in coming years as the labor orceparticipation rate or men has been alling while the participation rate

or women has been stable.The changing age composition o the workorce has led to em-

ployment shits in dierent directions. From 1983 to 2012, the em-

ployment share o workers age 55 and older in high-skill occupations


26/28


increased. This shit was related to the aging o the labor orce and the

delay in retirement o workers in highest demand those with higher

levels o education. In contrast, among workers ages 16 to 24 the larg-est increase was in the employment share o workers in low-skill oc-

cupations. Compared to the 1980s, younger people have been stayingin school longer and postponing their entry into the labor orce. Thesedevelopments have shited the composition o workers in the labor

orce and suggest that the retirement o the baby boom workers overthe next decade may reduce the supply o highly-skilled workers.

Job polarization has occurred consistently over the past three de-

cades, suggesting that it is an ongoing, structural phenomenon. Whilethe pace o job polarization accelerated during recessions, the continu-ous shits in the employment shares suggest that this phenomenon has

been neither a key driver o the business cycle nor the primary cause ojobless recoveries.


27/28


ENDNOTES

1For the purposes o this study, the data are restricted to workers ages 16 to

64 who are not sel-employed and are not employed in military or agricultural

occupations. To construct annual series, monthly observations are averaged or

each year. Similarly, quarterly series are constructed by taking averages o monthly

observations or each quarter.2This article ollows Acemoglu and Autor and Autor in categorizing occupa-

tions by skill level (high, middle and low).3Acemoglu writes, new technologies are not complementary to skills by

nature but by design, leading to a related but newer concept o directed techni-

cal change (Acemoglu 1998, 2002).

4In the CPS survey, wage data are collected or only two o the eight monthsin which a given household is surveyed. The sample size or wages, thereore, is

one-ourth the size o the ull data sample that is used or constructing employ-

ment shares. The analysis in this study uses wage data on workers weekly earn-

ings. As with the employment data, all observations within a given year are aver-

aged to create an annual series.5The decomposition is computed using the ollowing ormula:

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E-i s i s i i s

i

i s

i

i s

i

i i, ,2012

2012

, ,1983

1983

,2012

2012

, ,2012

,2012

, ,1983

,1983

, ,1983

,1983

,2012

2012

,1983

1983

=

+

where Erepresents employment, irepresents the industry sector, and srepresents

the skill level.6Changes due to job polarization primarily occurred within each gender

group rather than across gender groups. The emale employment share increased

only modestly during the period, rising rom 46 percent in 1983 to 48 percent

in 2012.7The most noticeable shit occurred ollowing the 2001 recession. However,

this episode appears to be a shit in the employment share away rom workers in

high-skill occupations toward low-skill occupations, which is not the type o shit

associated with job polarization.8The periods in which the economy was in recession between 1983 and 2012

were 1990Q4 to 1991Q1, 2001Q2 to 2001Q4, and 2008Q1 to 2009Q2 accord-

ing to the NBER business cycle dating committee. In total, this represents 11

quarters out o the total o 118 quarters in the sample rom 1983Q1 to 2012Q3.


28/28


REFERENCES

Acemoglu, Daron. 2002. Technical Change, Inequality and the Labor Market.Journal of Economic Literature, 40 (1): 7-72.

Acemoglu, Daron. 1998. Why Do New Technologies Complement Skills?Directed Technical Change and Wage Inequality. The Quarterly Journal ofEconomics, 113 (4): 1055-1089.

Acemoglu, Daron, and David H. Autor. 2011. Skills, Tasks and Technologies:Implications or Employment and Earnings in Orley Ashenelter and DavidCard (Eds.). Handbook of Labor Economics, 4B: 1043-1171. Elsevier.

Autor, David H. 2010. The Polarization o Job Opportunities in the U.S. LaborMarket: Implications or Employment and Earnings. Center or AmericanProgress and the Hamilton Project.

Autor, David H., Frank Levy, and Richard J. Murnane. 2003. The Skill Contento Recent Technological Change: An Empirical Exploration. TheQuarterlyJournal of Economics, 118 (4): 1279-1333.

Autor, David H., Lawrence F. Katz, and Alan B. Krueger. 1998. Computing In-equality: Have Computers Changed the Labor Market, The Quarterly Journalof Economics, 113 (4): 1169-1213.

Autor, David H., Lawrence F. Katz, Melissa S. Kearney. 2006. The Polarizationo the U.S. Labor Market.American Economic Review Papers and Proceedings,96 (2): 189-194.

Black, Sandra E., and Chinhui Juhn. 2000. The Rise o Female Proessionals:Are Women Responding to Skill Demand? The American Economic Review

90 (2): 450-455.Foote, Christopher, and Richard W. Ryan. 2012. Labor-Market Polarization

Over the Business Cycle. Federal Reserve Bank o Boston Public PolicyDiscussion Series No. 12-8.

Goldin, Claudia, and Lawrence F. Katz. 1998. The Origins o Technology-SkillComplementarity, The Quarterly Journal of Economics, 113 (3): 693-732.

Goos, Maarteen, Alan Manning, and Anna Salomons. 2011. Explaining JobPolarization: The Roles o Technology, Oshoring, and Institutions.University o Leuven, Center o Economic Studies Discussion Paper Series 11.34

Goos, Maarteen, and Alan Manning. 2007. Lousy and Lovely Jobs: The Rising

Polarization o Work in Britain. The Review of Economics and Statistics, 89(1): 118-133.Jaimovich, Nir, and Henry Siu. 2012. The Trend is the Cycle: Job Polarization

and Jobless Recoveries, NBER working paper 18334.Kimmel, Jean, and Thomas J. Kniesner. 1998. New Evidence on Labor Supply:

Employment Versus Hours Elasticities by Sex and Marital Status,Journal ofMonetary Economics, 42 (2): 289-301.

Krusell, Per, Lee E. Ohanian, Jose-Victor Rios-Rull, and Giovanni L. Violante.2000. Capital-Skill Complementarity and Inequality: A MacroeconomicAnalysis, Econometrica, 68 (5): 1029-1053.

Oldenski, Lindsay. 2012. O-Shoring and the Polarization o the Labor

Market, mimeo, Georgetown University.

13q1tuzemen willis 1

Documents