Full text of Monthly Labor Review : September 2020

View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

September 2020

COVID-19 recession is tougher on women
Eleni X. Karageorge
According to a new study, working women are experiencing the worst effects of the COVID-19 recession, unlike in
previous downturns, which hit working men the hardest. In “The Impact of COVID-19 on Gender
Equality” (National Bureau of Economic Research, Working Paper 26947, April 2020), researchers Titan Alon,
Matthias Doepke, Jane Olmstead-Rumsey, and Michèle Tertilt suggest that more women in the United States will
have lost their jobs because the industries they tend to work in have been harder hit by the effects of the
pandemic. The study points out two major reasons that the current recession is tougher for women.
First, the crisis has battered industry sectors in which women’s employment is more concentrated—restaurants
and other retail establishments, hospitality, and health care. This was not the case in past recessions, which
tended to hurt male-dominated industry sectors like manufacturing and construction more than other industries. In
past recessions, men have faced greater risk of unemployment than women, partly because of the gender
composition of different sectors of the economy. A larger fraction of employed men (46 percent) than employed
women (24 percent) work in construction; manufacturing; and trade, transportation, and utilities. These are
considered highly cyclical sectors that typically suffer during “normal” recessions. On the other hand, 40 percent of
all working women are employed in government and in health and education services compared with just 20
percent of working men.
Second, the coronavirus shutdowns have closed schools and daycare centers around the country, keeping kids at
home and making it even harder for parents (especially mothers who tend to provide the majority of childcare) to
keep working. Childcare poses an additional challenge to working mothers during the pandemic.
Working women are also at a greater disadvantage compared with working men in the current crisis because
fewer women have jobs that allow them to telecommute: 22 percent of female workers compared with 28 percent
of male workers. According to the researchers’ analysis of data from the American Time Use Survey from 2017
and 2018, single parents will face the greatest challenge. Only 20 percent of single parents reported being able to
telecommute compared with 40 percent of married people with children. In two-parent households where only one
parent works in the labor market, the stay-at-home parent, usually the mother, is likely to assume primary childcare
duties during coronavirus-related school closures. However, in 44 percent of married couples with children, both
spouses work full-time. Among these couples, mothers provide about 60 percent of childcare. Men perform 7.2
hours of childcare per week versus 10.3 hours for women.
The study also compared the current pandemic to World War II, which led to significant changes in American
family dynamics and gender norms, as millions of women joined the labor force for the first time. The huge influx of
women into the workforce during the Second World War was a change that did not revert once the war was over.
The comparison suggests that temporary changes to the distribution of labor between men and women may have
a lasting impact. As of now, millions of American fathers are taking more, and in some cases, primary responsibility
1

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

for child care. It is possible that the COVID-19 crisis may erode gender norms that currently lead to a lopsided
distribution of the division of labor in housework and childcare at home.
The current economic downturn resulting from the COVID-19 pandemic is disproportionately hurting women’s
employment, with ramifications that could be long lasting. The authors estimate that 15 million single mothers in
the United States will be the most severely affected, with little potential for receiving other sources of childcare and
a smaller likelihood of continuing to work during the crisis. However, the study points out that many businesses are
becoming much more aware of their employees’ childcare needs and have responded by adopting more flexible
work schedules and telecommuting options. The authors hope that by promoting flexible work arrangements and
making childcare obligations of both genders a priority, the crisis may reduce labor-market barriers in the long run.
Although the evidence suggests that women’s employment opportunities will suffer severely during the crisis, the
authors see cause for optimism over the longer term.

September 2020

Comparing characteristics and selected
expenditures of dual- and single-income
households with children
Using 2015–17 data from the Consumer Expenditure
Surveys, this article compares the food, transportation, and
education expenditures of dual- and single-income
households with children under age 18. The analysis finds
that these expenditures vary by both parental employment
status and children’s age.
The percentage of dual-income households with children
under age 18 has been on the rise since the 1960s,
surpassing the percentage of father-only-employed
households in the 1970s.1 This rise most likely reflects a
cultural shift involving women in the workforce. The female
labor force participation rate increased from 1960 onward,
peaking at 60 percent in 1999.2 Monitoring and analyzing
this trend is important, because the expenditure patterns of
dual-income households could differ from those of singleincome households, affecting the U.S. economy.
This article examines the characteristics and employmentstatus proportions of dual- and single-income (couple-led)

Julie Sullivan
sullivan.julie@bls.gov
Julie Sullivan is an economist in the Office of
Prices and Living Conditions, U.S. Bureau of
Labor Statistics.

households with children, comparing their expenditures on
food, transportation, childcare, and private education.
These expenditure categories are selected under the
assumption that working full time entails tradeoffs involving time for meal preparation, time for childrearing, and
commuting expenses. Using 2015–17 data from the Consumer Expenditure Surveys (CE), the analysis first
compares family characteristics (such as number of children) across the following three categories that capture the
employment status of parents in households with at least one full-time worker: “both full time,” “one full time, one
part time,” and “one full time, one not working.” The analysis then examines CE expenditure patterns, by children’s
age, within each employment category.

Data

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

The CE expenditure data are collected by the U.S. Census Bureau for the U.S. Bureau of Labor Statistics in two
component surveys: (1) the Interview Survey for major and/or recurring expenditure items and (2) the Diary Survey
for minor and/or frequently purchased expenditure items.3 (See appendix for more details about the data.) This
article uses internal microdata from both surveys.4 Data from the Interview Survey are used to compute
employment-status proportions and other family characteristics, as well as monthly household expenditures on
transportation, education, and childcare. Data from the Diary Survey are used to analyze weekly food
expenditures.
The present analysis uses a subset of CE data consisting of consumer units (similar to families) that reported
having a spouse and at least one child under age 18. This subset includes only married couples and their own
minor-age children (i.e., children under age 18); no other family members (e.g., grandparents) are included.5 As
noted earlier, three analysis groups are formed on the basis of the employment status of the couples,6 and the
analysis includes only couples who reported their employment status for the entire previous year.7 Full-time
employment is defined as working at least 35 hours a week, and part-time employment is defined as working 1 to
34 hours a week. Lastly, to control for expenditure differences between families with younger and older children,
the analysis breaks down the data by age of children in the household, forming three groups: households in which
all children are under age 6; households in which all children are ages 6 to 11; and households in which all children
are ages 12 to 17. To have large-enough sample sizes within each group, the analysis focuses on the 2015–17
period. The sample sizes are shown in table 1, by employment status and age of children, followed (in
parentheses) by the number of households represented nationally.
Table 1. Sample sizes and (in parentheses) number of represented households, 2015–17
Employment status
Age of children
One full time, one not working
All children under age 6
All children ages 6 to 11
All children ages 12 to 17

One full time, one part time

707 (1,215,385)
290 (461,944)
316 (540,438)

274 (480,165)
201 (340,279)
258 (429,887)

Both full time
1,047 (1,815,244)
665 (1,154,680)
914 (1,602,407)

Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

Employment-status proportions
CE data show that, among U.S. households, dual-income households have been a majority for at least the last two
decades. The percentage of dual-income households was fairly stable between 1998 and 2017, ranging from
52 to 58 percent. (See figure 1.) In this article, a dual-income household is defined as one in which one spouse
works full time and the other works at least part time. From 2007 to 2011, there was a steady decrease in the
percentage of dual-income households (from 58 to 53 percent for couples who had some kind of dual income), and
this decrease coincided with the Great Recession of 2007–09.8 In those years, the percentage of single-income
households increased, as did the percentage of households of other employment types (e.g., those in which both
spouses are not working or those in which one spouse is working part time).9

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

But what about families with children? Based on 2015–17 CE data (used in the rest of the analysis) for married
couples with children under age 18, the proportion of “one full time, one not working” households is 30 percent; the
proportion of “one full time, one part time” households is 14 percent; and the proportion of “both full time”
households is 52 percent. So, even among households with children, dual-income households make up two-thirds
(66 percent) of the total. This percentage is higher than that for the overall population (52 to 58 percent), partly
because retired couples (in which both spouses are considered not working) are more prevalent in the overall
population than among households with children.
Table 2 shows household proportions by both employment status and age of children. One can see that, as the
age of children increases from under age 6 to ages 6 to 11, the proportion of “one full time, one not working”
households decreases by 10 percentage points, and the proportion of “both full time” households increases by 8
percentage points.
Table 2. Percentage of households, by employment status and age of children, 2015–17
Employment status
Age of children
One full time, one not working
All children under age 6
All children ages 6 to 11
All children ages 12 to 17

33.4
23.0
19.6

Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

One full time, one part time
13.2
17.0
15.6

Both full time

Other

49.9
57.6
58.0

3.5
2.4
6.8

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Characteristics
Besides collecting expenditure data, the CE program collects demographic data from survey respondents. To get a
profile of single- and dual-income households, the analysis compares their average age, number of children, race,
income, and outlays across sample groups. This comparison is important because demographic characteristics
may affect household expenditures even within the same employment-status group. As shown in table 3, the age
of a household’s reference person10 varies little across employment types when the age range of children is held
constant. The average age of parents within each column follows a natural lifecycle function, increasing with
children’s age. As shown in table 4, the number of children in a household varies little across children age groups.
The largest difference (0.22) is again between “one full time, one not working” and “both full time” households, this
time for households in which all children are under age 6. For households in which all children are ages 12 to 17,
the difference is roughly halved (0.13).
Table 3. Average age of reference person in household, by employment status and age of children, 2015–
17
Employment status
Age of children
One full time, one not working
All children under age 6
All children ages 6 to 11
All children ages 12 to 17

One full time, one part time
33
41
48

Both full time
33
40
48

34
40
47

Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

Table 4. Average number of children in household, by employment status and age of children, 2015–17
Employment status
Age of children
One full time, one not working
All children under age 6
All children ages 6 to 11
All children ages 12 to 17

One full time, one part time

1.61
1.70
1.57

Both full time

1.55
1.64
1.48

1.39
1.49
1.44

Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

Figure 2 compares the distributions of households by employment status and number of children. (Again, this
comparison is for the sample restricted to households in which all children are under age 18.) As shown in the
figure, 61 percent of households in which both spouses work full time have just one child. This percentage
compares with about 53 percent of households with one spouse working full time and the other not working.
Therefore, a higher percentage of households in the latter group have two or more children. Comparing the
distributions shows that households with only one spouse working are more likely to have more children or that the
more children a household has, the less likely that both of its spouses will be working.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Lastly, table 5 shows household distributions by employment status and race (White, Black, and Asian).11 The race
designations are based on the race of the reference person, not necessarily of both spouses. Whites and Blacks
have similar distributions by household employment status. Asians exhibit a somewhat different pattern in that,
compared with Whites and Blacks, they have a higher percentage of “one full time, one not working” households
and, therefore, a lower percentage of households in the other two employment categories. A question worth further
investigation is whether families have a dual income because of necessity or personal preference. For example,
although the proportion of Black households in the “both full time” category is roughly the same as that for Whites,
the average income of these Black households is lower than the average income of their White counterparts. (See
table 6.) In fact, the average income of a “one full time, one not working” Black household is just under 60 percent
of that of a White or Asian household in the same employment category. However, because the CE do not ask
about the reasons for having a dual- or single-income status, the question cannot be answered with CE data.
Table 5. Percentage of households, by employment status and race of reference person, 2015–17
Employment status
Race
One full time, one not working
White
Black
Asian

One full time, one part time
27
23
39

Both full time
16
17
10

Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

58
61
51

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Table 6. Average annual household income, by employment status and race of reference person, 2015–17
Employment status
Race
One full time, one not working
White
Black
Asian

One full time, one part time

$79,374
46,950
81,798

Both full time

$109,184
67,397
76,182

$121,550
92,180
163,518

Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

Income and outlays
In considering income before taxes (hereafter referred to simply as “income”), one could reasonably expect that,
on average, dual-income households will have higher income than single-income households. But how much
higher? Table 7 shows the difference in income for the three analysis groups. As expected, compared with “one full
time, one not working” households, “both full time” and “one full time, one part time” households have higher
average annual incomes. However, the disparity across groups decreases with children’s age. For households in
which all children are under age 6, the income difference between “one full time, one not working” and “both full
time” households is $53,873. This difference drops to $19,718 for households in which all children are ages 12 to
17.
Table 7. Average annual total household income, by employment status and age of children, 2015–17
Employment status
Age of children
One full time, one not working
All children under age 6
All children ages 6 to 11
All children ages 12 to 17

$63,507
78,975
103,564

One full time, one part time
$92,319
109,403
113,192

Both full time
$117,380
117,023
123,282

Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

There is a noticeable shift in the age of employed spouses across employment-status groups. (See table 8.) This
shift is relevant to the analysis of average annual total income because, using age as a proxy for work experience,
one might expect that “one full time, one not working” households have a spouse with more work experience than
the spouses in “both full time” households. Among households in which all children are under age 6, “both full time”
households have a higher average age of employed spouses (34 years) than do “one full time, one not working”
households (33 years). However, among households in which all children are ages 12 to 17, “one full time, one not
working” households have the highest average age for the employed spouse. At the same time, although “both full
time” households have working spouses whose average age increases with children’s age, they are the only group
for which income does not rise as children’s age increases from less than 6 years to 6–11 years. Therefore, it is
unlikely that age, used as a proxy for work experience, is the only factor accounting for differences in total income.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Table 8. Age of full-time employed spouse and average age of both full-time employed spouses, by
employment status and age of children, 2015–17
Employment status
Age of children
One full time, one not working
All children under age 6
All children ages 6 to 11
All children ages 12 to 17

One full time, one part time

33
42
49

Both full time

34
41
49

34
40
47

Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

Table 9 presents total outlays,12 which serve as a proxy for permanent income.13 The figure shows that, compared
with income, total outlays vary less across analysis groups. It is interesting that, among households in which all
children are ages 6 to 11 or 12 to 17, “one full time, one part time” households have the highest average outlays.
The differences between the three groups decrease with children’s age. For households in which all children are
under age 6, the largest difference in outlays ($24,260) is between “one full time, one not working” and “both full
time” households. For households in which all children are ages 12 to 17, the largest difference ($11,976) is
between “one full time, one not working” and “one full time, one part time” households. In part, this decrease in
outlay disparities is presumably a function of the decrease in income disparities as children’s ages increase.
Another contributing factor is the decline in childcare expenditures for children older than age 6 (see childcare
expenditure analysis below).
Table 9. Average annual total household outlays, by employment status and age of children, 2015–17
Employment status
Age of children
One full time, one not working
All children under age 6
All children ages 6 to 11
All children ages 12 to 17

$57,760
77,524
83,104

One full time, one part time

Both full time

$73,764
94,940
95,080

$82,020
84,308
94,116

Note: In the Consumer Expenditure Interview Survey internal and microdata files, outlays are a quarterly amount. Because total income is often thought of as
an annual amount, the outlay variable was multiplied by 4.
Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

Food expenditures
Families face a tradeoff between spending time and spending money. Dual-income families forgo extra time on
meal preparation for the potential benefit of having higher total income, while single-income families forgo extra
income for the potential benefit of spending less money on childcare and food away from home. According to the
American Time Use Survey (ATUS), the time mothers spend on food preparation and cleanup is 0.8 hours per day
in “both full time” households and about twice that in households in which only one spouse (the father) works full
time.14 This section analyzes average weekly spending for food at home and food away from home for the three
types of employed households. Food at home is defined as food purchased from grocery or similar stores
(convenience stores, farmers’ markets, etc.), and food away from home is defined as food purchased at

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

restaurants, employer cafeterias, vending machines, or similar venues. The analysis examines weekly average
expenditures based on data from the CE Diary Survey, both because these data describe a person’s weekly
spending and because groceries and restaurant expenditures are often thought of in terms of weekly amounts.
Table 10 compares the food-at-home expenditures of the three analysis groups. The differences between the
groups are not statistically significant.15 This result may be partly due to the variety of frozen meals and prepared
foods that can be purchased at grocery stores. In fact, the data show that, compared with single-income
households, dual-income households spend consistently more on convenience foods (e.g., canned, preprepared,
or frozen foods).16 On average, “both full time” households spend $2.36 more per week on convenience foods
than do “one full time, one not working” households. This spending difference is statistically significant for
households in which all children are under age 6.
Table 10. Average weekly household expenditures for food at home, by employment status and age of
children, 2015–17
Employment status
Age of children
One full time, one not working
All children under age 6
All children ages 6 to 11
All children ages 12 to 17

One full time, one part time

$92.78 ($5.57)
113.40 (8.62)
124.04 (9.46)

Both full time

$110.20 ($11.23)
103.70 (10.34)
139.82 (9.98)

$104.04 ($5.59)
117.57 (7.06)
129.87 (6.63)

Note: Estimates represent mean expenditures. Standard errors are shown in parentheses.
Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

Furthermore, in terms of all food at home, the largest difference across groups is again observed for households in
which all children are under age 6. Among these households, “one full time, one part time” households spend
about $17 more per week, on average, than do “one full time, one not working” households. Similarly, the
difference between these two groups is $16 for households in which all children are ages 12 to 17.
Table 11 presents group comparisons for food-away-from-home weekly expenditures. Some of the differences in
this expenditure category are statistically significant. For example, among households in which all children are
under age 6, “one full time, one not working” households spend significantly less, on average, than do “one full
time, one part time” and “both full time” households.
Table 11. Average weekly household expenditures for food away from home, by employment status and
age of children, 2015–17
Employment status

t-values

Age of children
One full time, one not working (A) One full time, one part time (B)
All children under
age 6
All children ages 6
to 11

$53.89B,C ($5.38)

Both full time
(C)

$86.36A ($17.70) $82.89A ($6.36)

75.15 (12.15)

70.49 (12.54)

See footnotes at end of table.

t(A,B) t(A,C) t(B,C)

1.74 3.21 -0.18

94.00 (7.26) -0.27 1.30

1.62

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Table 11. Average weekly household expenditures for food away from home, by employment status and
age of children, 2015–17
Employment status

t-values

Age of children
One full time, one not working (A) One full time, one part time (B)
All children ages 12
to 17

89.98 (10.83)

92.81 (11.01)

Both full time
(C)
100.53 (6.72)

t(A,B) t(A,C) t(B,C)

0.21 0.89

0.59

Note: Superscripts indicate statistically significant differences between specific groups. For example, in column C, an “A” superscript indicates that the mean
for "both full time" households is significantly different from the mean for "one full time, one not working" households. Two superscripts in any column indicate
that the mean therein is significantly different from the means for the other two groups. Standard errors are shown in parentheses.
Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

Transportation expenditures
This section tests the hypothesis that, because of potentially higher commuting costs for two workers, dual-income
households would spend more on transportation than single-income households. This hypothesis is tested by
comparing monthly expenditures for public transportation (intercity bus, mass transit, and train) and gasoline, both
sourced from the CE Interview Survey.17
Contrary to the hypothesis, the results presented in table 12 show that the only significant difference in public
transportation expenditures is that between “one full time, one part time” and “both full time” households with
children ages 6 to 11. Intriguingly, for families with children ages 6 to 11, the difference does not appear to be due
to a difference in ownership of commuting vehicles (cars and trucks). According to data from the CE Interview
Survey, “one full time, one part time” households own about the same number of such vehicles (1.8, on average)
as do “both full time” households (1.9).
Table 12. Average monthly household expenditures for public transportation, by employment status and
age of children, 2015–17
Employment status

t-values

Age of children
One full time, one not working (A) One full time, one part time (B)
All children under
age 6
All children ages 6
to 11
All children ages 12
to 17

$6.68 ($1.93)

$11.64 ($5.17)

12.04 (5.04)

15.19C (3.73)

10.77 (3.98)

15.93 (5.19)

Both full time
(C)

t(A,B) t(A,C) t(B,C)

$9.41 ($1.65) 0.88

0.92 -0.43

7.85B (1.54) 0.46 -0.79 -1.85
10.78 (2.34) 0.68

0.00 -0.91

Note: Superscripts indicate statistically significant differences between specific groups. For example, in column C, a “B” superscript indicates that the mean for
"both full time" households is significantly different from the mean for "one full time, one part time" households. Standard errors are shown in parentheses.
Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Table 13 shows the results for gasoline expenditures, the second component of transportation expenditures. For
households in which all children are under age 6 or between the ages of 6 and 11, one finds a statistically
significant difference between “one full time, one not working” households and “both full time” households, with the
latter group spending more on gasoline than the former. This pattern shifts for households in which all children are
ages 12 to 17; here, “one full time, one not working” households spend the same, on average, as do “both full
time” households. Finally, regardless of household employment status, average expenditures on gasoline increase
with children’s age.
Table 13. Average monthly household expenditures for gasoline, by employment status and age of
children, 2015–17
Employment status

t-values

Age of children
One full time, one not working (A) One full time, one part time (B) Both full time (C) t(A,B) t(A,C) t(B,C)
All children under
age 6
All children ages 6
to 11
All children ages 12
to 17

$152.24C ($7.90)

$169.63 ($14.22)

$177.72A
($5.99)

1.07

2.23 0.56

177.62C (9.21)

193.70 (13.32)

200.83A (6.03)

0.94

2.07 0.53

220.97B (7.46)

201.46A (8.26)

217.21 (7.21) -1.76 -0.38 1.46

Note: Superscripts indicate statistically significant differences between specific groups. For example, in column C, an “A” superscript indicates that the mean
for "both full time" households is significantly different from the mean for "one full time, one not working" households. Standard errors are shown in
parentheses.
Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

Childcare expenditures
Many married couples with young children face the tradeoff between a second income and time spent with
children, particularly when considering the monthly cost of childcare. According to ATUS data, mothers in “both full
time” households spend 2.30 hours per day on caring for and helping household children. (See table 14.) This
figure is much higher (3.49 hours per day) for households in which the father is employed full time and the mother
is not employed. It is important to note that the ATUS data are categorized by the youngest, not the oldest, child in
the household. Therefore, the ATUS data are not directly comparable with CE data.
Table 14. Hours per day spent caring for and helping household children, by employment status and age
of children, 2015–17
Employment status
Age of children

Mother not employed, father

Mother employed part time, father

Both employed full

employed full time

time

Mother
Youngest child under age 6
Youngest child age 6 to 17

3.49
1.56

Father
1.13
0.52

Source: U.S. Bureau of Labor Statistics, American Time Use Survey.

Mother

Father
2.78
1.18

Mother
1.33
0.57

2.30
0.76

Father
1.54
0.50

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Table 15 shows the childcare expenses of households with children under age 6 and households with children
ages 6 to 11.18 These expenditures differ significantly—at the 90-percent confidence level—across employment
statuses within each children’s age category. For all employment statuses, the childcare expenditures of
households in which all children are ages 6 to 11 are substantially lower than the childcare expenditures of
households in which all children are under age 6. In fact, the average childcare expenditures for “one full time, one
not working” households in the former group drop to only $7, because children reaching school age no longer need
all-day daycare during the academic year.
Table 15. Average monthly childcare expenditures, by employment status and age of children, 2015–17
Employment status
Age of children
One full time, one not working
All children under age 6
All children ages 6 to 11

One full time, one part time

$60.51
6.74

Both full time

$252.85
33.11

$508.22
135.31

Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

Education expenditures
Unlike childcare expenditures, private education expenditures presumably reflect the personal preferences of
parents, because households have access to free public education. This section examines whether “both full time”
households spend more on private education than the other two employment-status groups. The comparison is
based on a variable that captures monthly expenditures on private tuition for elementary through high school.19
Because children under age 6 generally do not attend elementary school, the analysis is restricted to households
in which all children are ages 6 to 11 or ages 12 to 17.
Surprisingly, the spending differences across groups are not statistically significant. (See table 16.) Therefore, on
average, private school spending does not appear to differ between dual- and single-income households.
Table 16. Average monthly expenditures on private tuition for elementary through high school, by
employment status and age of children, 2015–17
Employment status
Age of children
One full time, one not working
All children ages 6 to 11
All children ages 12 to 17

$68.26
110.67

One full time, one part time
$229.41
158.37

Both full time
$75.54
101.53

Source: U.S. Bureau of Labor Statistics, Consumer Expenditure Surveys.

Summary and conclusion
In summary, the food, transportation, and education expenditures of dual- and single-income households depend
on the ages of household children. Childcare is the one expenditure category for which dual-income households
(“both full time” or “one full time, one part time” households) spend the most, regardless of children’s ages. As
expected, and again regardless of children’s ages, dual-income households have higher total incomes and total

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

outlays than single-income households. An interesting area for further research is the finding that “one full time,
one part time” households have the highest total outlays and the highest public transportation and private
education expenditures. This result may be due to these households being less time constrained than “both full
time” households and having higher incomes than “one full time, one not working” households. However, testing
this hypothesis would require a regression or another complex analysis that is beyond the scope of this article.
The present research is important for parents engaged in family planning or making career choices. By identifying
differences in the food, transportation, and childcare expenditures of dual- and single-income households, it can
help couples with children anticipate spending increases or decreases as they change their employment status or
as their children get older. In addition, the research can help retailers understand what goods and services are in
demand by dual-income families, which have represented most households in the last 20 years. Monitoring the
percentage of households in this category can facilitate market planning in the food, childcare, and private
education industries.

Appendix: About the data
CE data are collected by the U.S. Census Bureau for the U.S. Bureau of Labor Statistics in two component
surveys: the Diary Survey and the quarterly Interview Survey. The Diary Survey captures small expenditures, such
as those for groceries, personal care items, and housekeeping supplies, with respondents recording all purchases
over a 2-week period. The Interview Survey captures larger and/or recurring expenditures, such as those for
automobiles, major appliances, and rent and utilities. This survey is conducted every 3 months, for a total of four
in-person visits per year, asking respondents to recall items purchased in the previous 3 months.
According to the CE, “A consumer unit comprises either: (1) all members of a particular household who are related
by blood, marriage, adoption, or other legal arrangements; (2) a person living alone or sharing a household with
others or living as a roomer in a private home or lodging house or in permanent living quarters in a hotel or motel,
but who is financially independent; or (3) two or more persons living together who use their income to make joint
expenditure decisions. Financial independence is determined by the three major expense categories: housing,
food, and other living expenses. To be considered financially independent, at least two of the three major expense
categories have to be provided entirely, or in part, by the respondent.”20 Two roommates who share an apartment
but are otherwise financially independent are considered two consumer units within a household. Although some
married couples with young children may rent out a portion of their home, this article assumes that most of them
form a single consumer unit. For this reason, the discussion uses the terms “family” and “household”
interchangeably.
To be nationally representative, the data used in the analysis are weighted. Comparison statistics are derived from
a method called Balanced Repeated Replication, which estimates standard errors used in calculating t-statistics
and, hence, in significance testing. The data are divided into 43 groups, and each group is used to create a
randomly selected half-sample. From the resulting half-samples, 44 mean estimates are computed, and then the
standard error is calculated as the average of the difference between the half-sample estimates and the population
estimate.21
SUGGESTED CITATION

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Julie Sullivan, "Comparing characteristics and selected expenditures of dual- and single-income households with
children," Monthly Labor Review, U.S. Bureau of Labor Statistics, September 2020, https://doi.org/10.21916/mlr.
2020.19.
NOTES
1 Gretchen Livingston and Kim Parker, “8 facts about American dads,” Fact Tank: News in the Numbers (Washington, DC: Pew
Research Center, June 2019), https://www.pewresearch.org/fact-tank/2019/06/12/fathers-day-facts/.
2 Mitra Toossi and Teresa L. Morisi, “Women in the workforce before, during, and after the Great Recession,” Spotlight on Statistics
(U.S. Bureau of Labor Statistics, June 2017), https://www.bls.gov/spotlight/2017/women-in-the-workforce-before-during-and-after-thegreat-recession/pdf/women-in-the-workforce-before-during-and-after-the-great-recession.pdf.
3 For further information on CE data, see “Consumer expenditures and income,” Handbook of Methods (U.S. Bureau of Labor
Statistics), https://www.bls.gov/opub/hom/cex/home.htm.
4 Although the present analysis uses internal data, researchers can find BLS public-use microdata at https://www.bls.gov/cex/
pumd_data.htm.
5 For more information on consumer units and households, see appendix.
6 The CE data capture the main reason that a respondent did not work during the previous 12 months, such as unemployment,
retirement, or school attendance. For “one full time, one not working” households, this reason was “taking care of home/family.”
7 Each spouse must have reported working at least 50 weeks (full or part time) during the previous year.
8 According to the National Bureau of Economic Research, the recession began in December 2007 and ended in June 2009. See
“U.S. business cycle expansions and contractions” (Cambridge, MA: National Bureau of Economic Research), http://www.nber.org/
cycles.html.
9 According to labor force statistics from the Current Population Survey, the annual unemployment rate for people ages 16 and older
was 4.6 percent in 2007 and 8.9 percent in 2011, down from a peak of 9.6 percent in 2010. The annual data used here are not
seasonally adjusted and are obtained from https://data.bls.gov/PDQWeb/ln.
10 In the CE, the term “reference person” is defined as “the first member mentioned by the respondent when asked to ‘Start with the
name of the person or one of the persons who owns or rents the home.’” See “Consumer expenditures and income,” Handbook of
Methods (U.S. Bureau of Labor Statistics), p. 3.
11 People of other races, including multirace, are excluded from this analysis, because they constitute less than 1 percent of the
estimated population.
12 In the CE, expenditures on property include only mortgage interest, and expenditures on vehicles include the full value of the
purchased vehicle, whether or not the vehicle was financed. By contrast, outlays include both the principal and interest portions of
property mortgages and vehicle loans. The purchase price of vehicles bought outright and not financed also is included in outlays.
13 According to the “permanent income hypothesis,” first proposed by Milton Friedman in 1957, consumer expenditure decisions are
based not only on income received today but also on expectations of future income. See Friedman, “The permanent income
hypothesis,” in A theory of the consumption function (Cambridge, MA: National Bureau of Economic Research, 1957), pp. 20–37,
https://www.nber.org/chapters/c4405.pdf.
14 See table A-7A, “Time spent in primary activities by married mothers and fathers by employment status of self and spouse,
average for the combined years 2013–17, own household child under age 18,” American Time Use Survey (U.S. Bureau of Labor
Statistics), https://www.bls.gov/tus/tables/a7-1317.pdf.
15 The t-values for this analysis are derived by using the Balanced Repeated Replication method. For more information on this
method, see appendix and Kirk M. Wolter, Introduction to variance estimation, 2nd ed. (Chicago: Springer, 2007), p. 142.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

16 The items considered in this analysis include frozen vegetables, canned beans, canned corn, other canned vegetables, soup,
frozen meals, other frozen food, prepared salads, and miscellaneous prepared foods.
17 The CE Interview Survey collects quarterly data, but because transportation expenditures are often thought of in terms of monthly
amounts, the variables used in calculating public transportation and gas expenditures were divided by 3 for this analysis.
18 Like transportation expenditures, childcare expenditures are often thought of in terms of monthly amounts, so the variables used in
calculating childcare expenditures were divided by 3 for this analysis. Children ages 12 and 17 usually do not need “childcare,” which,
according to the CE, includes babysitting, daycare, nursery, and preschool; the childcare expenditures for children in this age group
are at or near $0 for all three household employment statuses.
19 Private school tuition is also collected in the CE Interview Survey. Although tuition is often thought of in terms of annual amounts,
households often budget in terms of monthly amounts. The variables used in calculating education expenditures were divided by 3 for
this analysis.
20 See “Glossary,” Consumer Expenditure Surveys (U.S. Bureau of Labor Statistics), https://www.bls.gov/cex/csxgloss.htm.
21 For further information on this methodology, see “Consumer expenditures and income,” Handbook of Methods (U.S. Bureau of
Labor Statistics), section “Calculation precision,” https://www.bls.gov/opub/hom/cex/pdf/cex.pdf. For an explanation of the Balanced
Repeated Replication method, see “Balanced Repeated Replication (BRR) method,” SAS/STAT(R) 9.2 user’s guide, 2nd ed. (SAS),
https://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#statug_surveymeans_a0000000225.htm.

RELATED CONTENT

Related Articles
Coworking couples and the similar jobs of dual-earner households, Monthly Labor Review, November 2019.
Tradeoffs in the expenditure patterns of families with children, Beyond the Numbers, July 2018.
Fun facts about Millennials: comparing expenditure patterns from the latest through the Greatest generation, Monthly Labor Review,
March 2018.
Program participation and spending patterns of families receiving government means-tested assistance, Monthly Labor Review,
January 2018.

Related Subjects
Income

Education and training

Family issues

Consumer expenditures

Demographics

Child care

Part time work

Social issues

September 2020

COVID-19, educational attainment, and the impact
on American workers
Cody Parkinson
The coronavirus disease 2019 (COVID-19) has disrupted the U.S. labor market in many ways. Early data indicate
that American workers without a college degree have experienced the most severe impact. In “The unequal impact
of COVID-19: why education matters” (FRBSF Economic Letter, Federal Reserve Bank of San Francisco, June 29,
2020), Mary C. Daly, Shelby R. Buckman, and Lily M. Seitelman examine how American workers with different
levels of education have fared since the start of the pandemic.
With employment levels falling by more than 20 million from March through May 2020 and the unemployment rate
rising to its highest level since 1938, many people are feeling the economic impact of COVID-19. As of May 2020,
53 percent of Americans were working, compared with 61 percent at the start of the year. While people are losing
their jobs, many are also leaving the labor force. Other factors that contributed to the disproportionate impact
include occupation, industry, social distancing, and shelter-in-place measures.
During the period of expansion before COVID-19, many long-standing economic gaps between more and less
advantaged groups narrowed. For example, the difference between the unemployment rate for those with a high
school diploma or less and a bachelor’s degree or more was 2.2 percentage points. The gap increased to 8.8
percentage points in May 2020. The unemployment rate for people with a high school diploma or less rose more
than 12.0 percentage points between February and May 2020. The rate rose 5.5 percentage points during the
same period for those with a bachelor’s degree or more. People with a high school diploma or less also
experienced a larger decline in labor force participation, decreasing 4.0 percentage points to 51.8 percent from
February to May 2020. Those with a bachelor’s degree or more had a labor force participation rate of 71.9 percent
at the end of the period, declining 1.2 percentage points. The proportion of job loss relative to working-age
population share was also higher among those with a high school diploma or less.
The authors examined other factors potentially contributing to the disproportionate impact of COVID-19 on workers
with different levels of education. Approximately 65 percent of workers with a bachelor’s degree or more
teleworked during COVID-19. In contrast, 22 percent of workers with a high school diploma or less teleworked,
reflecting structural job vulnerabilities to mandated changes in work environments for workers with lower levels of
education. Data also show that workers with higher levels of education are more likely to hold jobs that involve less
interpersonal contact. Many workers at nonessential businesses—hospitality, personal care, and restaurants—
were affected by the pandemic more than essential businesses because of social distancing and other health
precautions.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Daly, Buckman, and Seitelman close by noting that these discrepancies were not created during this COVID-19
pandemic, rather the inequalities were deepened. In addition, they note that access to education is not equal and
that greater access to higher education may ensure better economic resiliency in the future for the country and its
people.

September 2020

Exploring changes in real average hourly
earnings, June 2009 to December 2019
This article examines trends in real average hourly earnings
(1982–84 dollars) for all employees from June 2009, the
trough of the 2007–09 recession, to December 2019. It
looks at real earnings at the total private and major industry
levels, with more detailed analysis for select industries. The
article analyzes what drove the postrecession growth in real
hourly earnings. In particular, it identifies which industries
contributed the most to overall earnings growth during the
period.
The U.S. Bureau of Labor Statistics produces several
measures of pay and benefits.[1] This article focuses on
one of the timeliest measures: real average hourly earnings
(1982–84 dollars) from the Current Employment Statistics
(CES) survey, also known as the establishment or payroll

Lawrence Doppelt

survey.[2] Every month, the CES survey produces detailed
estimates of industry employment, hours, and earnings of
workers on nonfarm payrolls.[3] The economy added

Lawrence Doppelt is an economist formerly in the
Office of Employment and Unemployment
Statistics, U.S. Bureau of Labor Statistics.

millions of jobs during the economic expansion that began
Shane Haley
haley.shane@bls.gov

in June 2009, but whether or not these are “good” jobs
remains an open question.
In this article, we examine trends in real (inflation-adjusted)
average hourly earnings of all employees from June 2009,

Shane Haley is an economist in the Office of
Employment and Unemployment Statistics, U.S.
Bureau of Labor Statistics.

the trough of the 2007–09 recession as determined by the
National Bureau of Economic Research, to December
2019.[4] We look at hourly earnings at the total private and
major industry levels, with a more detailed analysis for select industries.[5] Our goal is to analyze what drove the
postrecession growth in real hourly earnings. That is, we identify which industries contributed the most—or the
least—to earnings growth at the highest level and how these sectors affected the labor market as a whole.

Basis and scope

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Average hourly earnings from the CES survey are a measure of gross payrolls divided by total hours for which
employees receive pay—including sick pay or vacation pay—during the pay period that includes the 12th of the
month. They do not represent employers’ total compensation costs because they exclude items such as employee
benefits, irregular bonuses and commissions, retroactive payments, and the employer’s share of payroll taxes.[6]
How have earnings changed since the end of the 2007–09 recession? Figure 1 illustrates the rise in total private
real average hourly earnings, with a notable uptick in earnings beginning in 2014. Real hourly earnings rose by 68
cents, from $10.30 in June 2009 to $10.98 in December 2019, for a 6.6-percent total increase. Real earnings also
increased over the 2009–19 period for every major industry group.

Several key components factor into total private earnings growth. Nominal average hourly earnings for all
employees from the CES survey are calculated as follows: Aggregate weekly payrolls of all employees divided
by aggregate weekly hours of all employees. This formula represents earnings at the nominal level. However, in
this article, we use “real” or inflation-adjusted hourly earnings to compare earnings in 2009 with earnings in 2019
by removing the inflation factor. Real earnings are a better reflection of purchasing power and provide a more
accurate comparison of earnings at different points in time. Using current-dollar values—as opposed to real or
“constant-dollar” values—would show large increases in earnings across all industries, which would overstate
earnings growth by not accounting for inflation.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

We use a deflator to calculate real earnings from the nominal values. For all employees, the deflator is the
Consumer Price Index for All Urban Consumers (CPI-U).[7] Therefore, real average hourly earnings for all
employees are calculated as follows: Real aggregate weekly payrolls of all employees divided by aggregate
weekly hours of all employees.

Real earnings as a function of the payroll-to-hours relationship
At its most basic level, the growth in real earnings stems from real aggregate payrolls increasing at a faster rate
than total hours worked. Therefore, mathematically, both the numerator and the denominator in the real earnings
formula play a key role in the direction of the data series. That is, if the numerator increases more quickly than the
denominator, then earnings must increase. The opposite holds true as well—if the denominator increases at a
faster rate than the numerator, then earnings must decrease. Figure 2 illustrates this relationship clearly.

Payrolls and hours increased at almost the same rate from June 2009 until the middle of 2014, with real earnings
growth hovering around 0 percent. At some points during the early part of the recovery, aggregate hours
(denominator) were even growing slightly faster than aggregate payrolls (numerator), yielding a decline in real
earnings. When payrolls began to increase faster than hours, as we see beginning in the second half of 2014, real
earnings grew and continued to do so through 2019.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Total private real average hourly earnings have been rising since 2014, so which industries are driving this growth?
Table 1 shows the percentage change in real total payrolls and aggregate weekly hours at the major industry level
over the June 2009–December 2019 period. It thus shows the payroll-to-hours relationship, with the most relevant
information being the difference between payroll and hours growth. If payrolls increase quickly and hours increase
at a similar rate, then they will cancel each other out and real earnings growth will be stagnant. From 2009 to 2019,
payrolls increased more than hours in all major industries, leading to an increase in total private real average
hourly earnings over the period.
Table 1. Percentage change in real aggregate weekly payrolls and aggregate weekly hours for all
employees, by major industry, June 2009–December 2019
Industry
Total private
Mining and logging
Construction
Manufacturing
Durable goods
Nondurable goods
Trade, transportation,
and utilities
Wholesale trade
Retail trade
Transportation and
warehousing
Utilities
Information
Financial activities
Professional and
business services
Education and health
services
Health care and social
assistance
Leisure and hospitality
Other services

Change in aggregate

Change in aggregate weekly

Difference, in percentage

weekly payrolls

hours

points

29.4%
19.8
37.3
16.4
18.8
12.1

21.4%
14.2
31.8
14.5
18.4
8.7

8.0
5.6
5.5
1.8
0.4
3.4

19.1

12.3

6.8

14.9
14.6

10.5
5.7

4.4
8.9

38.2

35.4

2.7

9.7
23.8
32.7

2.7
2.8
15.8

7.1
21.0
16.9

41.3

34.4

6.8

32.1

26.3

5.9

32.3

26.5

5.7

40.0
21.5

29.9
12.1

10.1
9.5

Note: Payroll data are adjusted for inflation using the Consumer Price Index for All Urban Consumers (CPI-U). Data are seasonally adjusted.
Source: U.S. Bureau of Labor Statistics.

Table 2 shows the percentage change in real earnings from June 2009 to December 2019. As can be seen in the
table, real earnings increased in all of the major industry groups, which indicates that payrolls increased at a faster
rate than hours for each of the groups during the postrecession period. Transportation and warehousing
experienced very high payroll and hours growth over the period (as shown in table 1), but because pay and hours
increased at almost the same rate, real earnings rose by only 2.0 percent.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Table 2. Real average hourly earnings (1982–84 dollars) and percentage change, all employees, by major
industry, June 2009–December 2019
Industry
Total private
Mining and logging
Construction
Manufacturing
Durable goods
Nondurable goods
Trade, transportation, and
utilities
Wholesale trade
Retail trade
Transportation and
warehousing
Utilities
Information
Financial activities
Professional and business
services
Education and health
services
Health care and social
assistance
Leisure and hospitality
Other services

Real average hourly earnings,

Percentage

June 2009

December 2019

change

$10.30
12.75
11.57
10.72
11.42
9.60

$10.98
13.38
12.05
10.89
11.46
9.90

6.6%
4.9
4.1
1.6
0.4
3.1

8.94

9.48

6.0

11.76
7.15

12.23
7.75

4.0
8.4

9.45

9.64

2.0

15.26
13.68
12.33

16.31
16.47
14.13

6.9
20.4
14.6

12.60

13.24

5.1

10.30

10.78

4.7

10.39

10.86

4.5

6.02
9.12

6.49
9.89

7.8
8.4

Note: Data are adjusted for inflation using the Consumer Price Index for All Urban Consumers (CPI-U). Data are seasonally adjusted.
Source: U.S. Bureau of Labor Statistics.

On the other end of the spectrum, the rate of payroll growth in the information industry (23.8 percent) was about
average relative to other industries, yet real earnings increased by a substantial 20.4 percent, as aggregate hours
in the industry increased by only 2.8 percent. (See table 1.) As a result, information had the largest percentage
change in real earnings of all major industries over the period; at $16.47 per hour, information also had the highest
earnings rate, surpassing that of utilities ($16.31), which had the highest rate ($15.26) in June 2009. (See table 2.)
This large earnings growth occurred because information had the largest difference between payroll growth (23.8
percent) and hours growth (2.8 percent). Hence, although information had relatively little increase in total hours
worked between 2009 and 2019, establishments in the industry were paying their employees nearly 24 percent
more by the end of the period.

Employment
At first glance, information, financial activities, and utilities industries might appear to have performed the best
during the postrecession period, because they make up the top three industries in terms of real earnings growth
and they are the highest paying industries overall. On the basis of these data alone, one might assume that the
three industries together represent one of the driving forces of the strong economy during the 2009–19 period.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

This assumption ignores a major factor, though—employment. Table 3 shows the number of employees added to
each major industry from June 2009 to December 2019. Over that period, total private employment increased by
about 20.9 million, with some industries adding many more jobs than others.
Table 3. Change in employment and percentage change, all employees, by major industry, June 2009–
December 2019
Industry

Change in employment

Total private
Mining and logging
Construction
Manufacturing
Durable goods
Nondurable goods
Trade, transportation, and utilities
Wholesale trade
Retail trade
Transportation and warehousing
Utilities
Information
Financial activities
Professional and business services
Education and health services
Health care and social assistance
Leisure and hospitality
Other services

Percentage change

20,887,000
29,000
1,545,000
1,140,000
882,000
258,000
2,974,000
420,000
1,131,000
1,436,000
–12,000
87,000
994,000
5,006,000
4,851,000
4,133,000
3,708,000
553,000

19.3%
4.2
25.7
9.7
12.3
5.7
12.0
7.6
7.8
34.0
–2.2
3.1
12.7
30.3
24.7
25.0
28.4
10.3

Note: Data are seasonally adjusted.
Source: U.S. Bureau of Labor Statistics.

The professional and business services industry added the most jobs over the 2009–19 period—about 5 million, or
nearly 24 percent of the total jobs gained—with education and health services close behind. Interestingly,
employment in utilities actually declined over the 10-year span. Although utilities had the highest real average
hourly earnings level as well as positive earnings growth over the period, the industry had fewer employees in
December 2019 than it had in June 2009.
Therefore, it is difficult to argue that an industry is driving real earnings growth when it has been losing workers.
One job added to the economy increases both payrolls and hours. In order to boost total private earnings, that job
must pay enough so that the marginal added payroll-to-hours ratio exceeds the total private average. In other
words, employment growth in industries such as utilities, information, and professional and business services will
more likely put upward pressure on the total private earnings average because these industries have a higher
earnings rate than the total private average. One job added to leisure and hospitality, on the other hand, will likely
decrease the total private average because the industry’s hourly earnings rate is lower than the total private rate.
The opposite holds true as well. Although adding one job to utilities would put upward pressure on total private
earnings, eliminating one job in the industry would have a negative impact on total private earnings—which is what
actually happened in the industry, as utilities employment declined slightly (−12,000) over the period. In this regard,
employment acts as a weight on that upward or downward pressure on total private earnings. Because pressure
6

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

on total private earnings is affected by scale, a small industry such as utilities does not have as much impact on
overall earnings as a large industry such as professional business services.
For example, we can compare the information industry, which is similar to utilities in terms of employment size and
earnings level, to professional and business services. Although information had a higher real earnings level in
December 2019 ($16.47) than professional and business services ($13.24), the latter added 5 million jobs to the
economy, while the former added only 87,000. Generally speaking, because information had a higher earnings
rate than professional and business services, adding one job to that industry would increase real earnings at the
total private level more than adding one job to professional and business services.
A perfectly efficient labor market never exists in reality. Both industries contributed in a positive way to total private
earnings. But because professional and business services added many more jobs than information, it had a
stronger positive effect on the labor market than information did over the 10-year period. To summarize this effect,
table 4 displays the ratio of real earnings in each industry group to total private earnings. A ratio above 1 signifies
that the industry had a higher real earnings level than the total private average and thus any jobs added would put
upward pressure on the total private earnings average. Similarly, ratios below 1 indicate that added jobs would put
downward pressure on the total private average. Industries with decreasing employment have the opposite effect:
a ratio greater than 1 will negatively affect the total private earnings level, while a ratio of less than 1 will positively
affect the total private earnings level.
Table 4. Ratio of real average hourly earnings in specific industries to total private level, all employees,
June 2009 and December 2019
Industry

June 2009

December 2019
1.00
1.24
1.12
1.04
1.11
0.93
0.87
1.14
0.69
0.92
1.48
1.33
1.20
1.22
1.00
1.01
0.58
0.89

Source: U.S. Bureau of Labor Statistics.

The dynamic relationships between the earnings levels in individual industries and the total private level are
evident in table 4. Some sectors, such as information, utilities, and financial activities, experienced high wage
growth relative to the total private average, indicated by a higher ratio in December 2019 than in June 2009.
7

1.00
1.22
1.10
0.99
1.04
0.90
0.86
1.11
0.71
0.88
1.49
1.50
1.29
1.21
0.98
0.99
0.59
0.90

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Others, such as durable goods manufacturing and transportation and warehousing, experienced diminished
relative wage growth, represented by their lower ratios. The ratios in manufacturing and education and health
services went below 1 over the period, meaning that, on average, adding one job would likely have exerted upward
pressure in 2009 but downward pressure in 2019.
As mentioned previously, employment acts as a weight for both payrolls and hours and therefore indirectly affects
total private earnings. Differences in weekly hours across industries complicate this issue further. Some industries
carry more part-time workers, while others are more likely to have overtime. Also, some industries pay higher
wages, on average, which can be attributed to a variety of factors not limited to productivity, education, labor
demand, and so on.
To help us visualize the weights by industry, figure 3 illustrates what happened over the 2007–09 period in the
major industry groups. The sizes of the bubbles in figure 3 represent the total employment in each of the
industries in December 2019, the y-axis represents real average hourly earnings in December 2019, and the x-axis
represents the number of jobs added from June 2009 to December 2019. The horizontal dotted line represents the
total private real average hourly earnings level of $10.98 in December 2019, and the vertical dotted line represents
the average added employment across all 14 industry groups over the period, which is about 1.5 million jobs. From
there, we assigned quadrants as follows: Quadrant I contains the industries with above-average earnings and
below-average employment gains, quadrant II contains the industries with above-average earnings and aboveaverage employment gains, quadrant III contains the industries with below-average earnings and above-average
employment gains, and quadrant IV contains the industries with below-average earnings and below-average
employment gains.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

As long as the industry is in quadrants I or II with positive employment, it puts upward pressure on total private
earnings. Industries in quadrant II are more impactful than those in quadrant I in that regard because they had
above-average earnings and added an above-average amount of employment. Industries in quadrants III and IV
negatively affect the total private earnings level when they add employment. Therefore, the leisure and hospitality
industry, which is in quadrant III, puts strong downward pressure on average earnings at the total private level.
However, recall from table 2 that leisure and hospitality experienced a 7.8-percent increase in real average hourly
earnings over the 2009–19 period. Although the industry still put downward pressure on total private earnings
growth, it is not as much as it was in the earlier part of the study period. From Figure 3, we see that professional
and business services was the most impactful major industry for increases in total private real earnings because it

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

had higher-than-average earnings and added a very large amount of employment over the period. The education
and health services industry had a relatively neutral impact because its earnings were roughly in line with the total
private average.

Detailed industry analysis
Because employment acts as a weight for both the numerator and the denominator in the average hourly earnings
equation, the results can sometimes be surprising. In a 2018 article, Angela Clinton discusses the math behind
counterintuitive scenarios such as when earnings increase in the component industries but decrease at the more
aggregate level.[8] The interactions of payrolls, hours, and employment occasionally lead to statistical anomalies
such as Simpson’s Paradox.[9] Thus, it can be difficult to determine which industries are performing the best,
which are having the heaviest impact on total private trends, and the reasons why.
To understand the total private trends, we analyzed the major industry groups. A natural response to viewing the
major industries is to ask, Why are some sectors doing well and others lagging behind? We can attempt to answer
this question by examining the data at a more detailed industry level. As an example, we take a more detailed look
at the information sector.
As shown in table 2, the information industry posted the highest growth in real average hourly earnings over the
2009–19 period. We saw that payrolls increased modestly, but hours and employment had only minor gains. So
what is happening in the information industry that is causing its earnings to increase more than those of other
industries? The hours and employment relationship makes sense, because small gains in employment tend to lead
to small increases in total hours worked. From an economist’s perspective, if labor demand is high and labor
supply is low, then wages could increase while employment remains flat.
The information industry has six component industries: publishing industries, except Internet; motion picture and
sound recording industries; broadcasting, except Internet; telecommunications; data processing, hosting and
related services; and other information services. Table 5 shows the percentage change in real hourly earnings for
the information industry and each of its components over the period from June 2009 to December 2019. Looking at
the table, we can see why information is the leading sector in terms of earnings growth—each of its components
shows above-average earnings gains. All components except motion picture and sound recording industries and
telecommunications experienced earnings growth rates of more than 20 percent.
Table 5. Percentage change in real average hourly earnings (1982–84 dollars) for the information industry
and its components, all employees, June 2009–December 2019
Industry
Information
Publishing industries, except Internet
Motion picture and sound recording
industries
Broadcasting, except Internet
Telecommunications
Data processing, hosting, and related
services

June 2009

December 2019

Percentage change

$13.68
15.03

$16.47
18.21

20.4%
21.2

13.04

15.05

15.4

12.81
13.07

15.95
15.19

24.5
16.2

13.70

16.73

22.1

See footnotes at end of table.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Table 5. Percentage change in real average hourly earnings (1982–84 dollars) for the information industry
and its components, all employees, June 2009–December 2019
Industry

June 2009

Other information services

December 2019

14.04

Percentage change

16.96

20.8

Note: Data are adjusted for inflation using the Consumer Price Index for All Urban Consumers (CPI-U). Data are seasonally adjusted.
Source: U.S. Bureau of Labor Statistics.

The largest component within the information industry—publishing industries, except Internet—provides an
interesting story. The industry consists of two component industries, newspaper, book, and directory publishers;
and software publishers. Table 6, shows that in June 2009, two-thirds of employment in publishing industries,
except Internet, was in newspaper, book, and directory publishers, with 536,800 employees and real average
hourly earnings of $11.66. The software publishing industry had only 257,800 employees in June 2009, but its
hourly earnings level was more impressive, at $21.40.
Table 6. Employment and real average hourly earnings (1982–84 dollars) in publishing industries, except
Internet, all employees, June 2009 and December 2019
June 2009

December 2019

Industry
Employment
Publishing industries, except Internet
Newspaper, book, and directory
publishers
Software publishers

Average hourly earnings

Employment

Average hourly earnings

794,700

$15.03

764,400

$18.21

536,800

11.66

287,700

11.75

257,800

21.40

477,100

21.85

Note: Data are adjusted for inflation using the Consumer Price Index for All Urban Consumers (CPI-U). Data are seasonally adjusted.
Source: U.S. Bureau of Labor Statistics.

By the end of 2019, employment in newspaper, book, and directory publishers had declined to 287,700 and hourly
earnings had increased to $11.75, whereas employment in software publishers had expanded to 477,100 and
earnings had increased to $21.85. Those changes represent an 85-percent increase in employment for software
publishers and a 46-percent decrease in employment for newspaper, book, and directory publishers. As noted
previously, increasing employment in industries with above-average earnings increases total private earnings,
while declining employment in industries with below-average earnings also increases total private earnings. Thus,
the increase in real earnings in publishing industries, except Internet, resulted from its lower paying component
industry losing employment and its higher paying component industry gaining employment.
The publishing industry itself has shifted from a paper-based industry to a largely electronic industry, shedding jobs
in the former and gaining them in the latter. The diverging employment trends between software publishers and
newspaper, book, and directory publishers are not visible at the aggregate level of publishing industries, except
Internet, let alone at the information-sector level. The trends at the detailed level remind us that these are what
drive the higher level industry trends. Major industries are not monolithic entities in which employment levels move
in one direction. Some specific industries, such as software publishers, had substantial increases in employment

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

over the 2009–19 period, while others, such as newspaper, book, and directory publishers, had substantial
declines.

Conclusion
Although the question of which industries are driving real average hourly earnings growth is nuanced, other factors
are clear. Adding employment to a relatively high-earning industry will drive overall earnings up; in that sense,
professional and business services was the most successful industry because it added about 5 million jobs, or
nearly a fourth of the 20.9 million total jobs added since the end of the 2007–09 recession. Employment in other
industries, such as utilities and information, changed little, while real earnings grew rapidly. Even relatively lowpaying industries, such as leisure and hospitality, had improvement in their real earnings. Substantial employment
gains in leisure and hospitality lowered total private average earnings but not as much as occurred during the last
recession.
The monthly CES survey data during the period of expansion from June 2009 to December 2019 raise the
question, Is the economy adding good jobs? That question has always been difficult to answer. Although every
major industry experienced real earnings growth over the period, detailed industry analysis reveals more nuanced
trends. The real earnings gap between certain industries has grown as well, as high-earning industries such as
information and utilities have seen their earnings increase at a faster rate than low-earning industries. As this
article demonstrates, real average hourly earnings can be a complex statistic.
SUGGESTED CITATION

Lawrence Doppelt and Shane Haley, "Exploring changes in real average hourly earnings, June 2009 to December
2019," Monthly Labor Review, U.S. Bureau of Labor Statistics, September 2020, https://doi.org/10.21916/mlr.
2020.20.
NOTES
1 The U.S. Bureau of Labor Statistics (BLS) has 12 surveys or programs that provide information on pay and benefits. For more
information, see “Overview of BLS statistics on pay and benefits,” https://www.bls.gov/bls/wages.htm.
2 For the latest issue of the real earnings news release, see https://www.bls.gov/news.release/realer.toc.htm. For technical
information about the real earnings data, see “Real earnings technical note” (part of the news release), https://www.bls.gov/
news.release/realer.tn.htm.
3 The Current Employment Statistics (CES) program surveys about 145,000 private businesses and government establishments each
month, representing approximately 697,000 individual worksites. For more information, see “Current Employment Statistics—CES
(national),” https://www.bls.gov/ces.
4 The starting and ending dates of recessions are determined by the National Bureau of Economic Research (NBER). NBER
determined that the peak of the most recent expansion occurred in February 2020. For more information, see “U.S. business cycle
expansions and contractions” (National Bureau of Economic Research, June 8, 2020), http://www.nber.org/cycles.html.
5 Throughout this article, “average hourly earnings,” “earnings,” and “hourly earnings” are used interchangeably. All data discussed in
this article are seasonally adjusted and adjusted for inflation using the Consumer Price Index for All Urban Consumers (CPI-U).
6 For more information about CES survey earnings and other concepts, see “Current Employment Statistics—CES (national):
technical notes for the Current Employment Statistics survey,” https://www.bls.gov/web/empsit/cestn.htm.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

7 BLS does not produce real aggregate weekly payrolls of all employees as a distinct time series; however, historical real aggregate
payrolls can be derived from other existing data series.
8 Angela Clinton, “An average mystery in hours and earnings data entails a weighty explanation,” Beyond the Numbers: Employment
and Unemployment, vol. 7, no. 9, June 2018, https://www.bls.gov/opub/btn/volume-7/mystery-in-average-of-hours-and-earnings.htm.
9 For more information, see “Simpson’s Paradox,” Stanford Encyclopedia of Philosophy (Stanford University, 2020), https://
plato.stanford.edu/entries/paradox-simpson/.

RELATED CONTENT

Related Articles
Demographics, earnings, and family characteristics of workers in sectors initially affected by COVID-19 shutdowns, Monthly Labor
Review, June 2020
Employment expansion continued in 2019, but growth slowed in several industries, Monthly Labor Review, April 2020
The relative weakness in earnings of production workers in manufacturing,
1990–2018, Monthly Labor Review, December 2019

Related Subjects
Earnings and wages

Consumer price index

Pay trends

industry

Expansions

Inflation

September 2020

How did the 2005 hurricanes affect individuals’
long-term earnings?
Jeffrey A. Groen, Mark J. Kutzbach, Anne E. Polivka
Major disasters, such as hurricanes, typically reduce labor market activity in the affected area during the
immediate aftermath. The physical damage and evacuations caused by these disasters can stop some businesses
from operating and some individuals from working or getting to their jobs. As a result, individuals who reside in a
disaster area often experience declines in their employment and earnings in the short term. But what happens over
the long term?
In “Storms and jobs: the effect of hurricanes on individuals’ employment and earnings over the long term” (Journal
of Labor Economics, July 2020), we examine the effects of Hurricanes Katrina and Rita, which devastated the U.S.
Gulf Coast in 2005. We estimate effects on the earnings of affected workers over the short and long term. To
explain the pace of the recovery of a worker’s earnings, we demonstrate the importance of two factors: damage to
the worker’s home and workplace and the worker’s prestorm industry of employment. We combine Federal
Emergency Management Agency damage data with U.S. Census Bureau data from household surveys and
longitudinal administrative data on jobs and places of residence. From these combined individual-level data, we
construct treatment and control samples, which we compare to estimate the effect of the storms on earnings. Our
treatment sample consists of individuals who resided (at the time of the storms) in storm-affected areas in four
states (Louisiana, Mississippi, Texas, and Alabama). Our control sample consists of individuals who resided in
similar unaffected areas elsewhere in the United States. The populations and economies of these areas were
similar to those of the storm-affected areas. In our job-level data, which are compiled from the Longitudinal
Employer-Household Dynamics program, we track a worker’s quarterly earnings from 2 years before the storms to
7 years after the storms.
We find that over the first year after the storms, the earnings of individuals affected by the hurricane were reduced.
The earnings losses, which were due primarily to job loss, reflect various short-term disruptions caused by the
hurricanes. Individuals whose home or workplace was damaged experienced larger earnings losses in the short
term. These losses may be attributed to workers moving out of the affected area or to businesses closing down. In
the medium and long term, affected individuals experienced earnings gains, primarily because of earnings gains
within employment. Over the entire poststorm period covered by our data, the storms led to a net increase in the
average quarterly earnings of affected individuals. Although earnings increased overall, they varied widely across
individuals, depending on their industry and the degree of damage to their home or workplace.
We show that the long-term earnings gains of affected individuals were the result of differences in local labor
market dynamics between the affected areas and the control areas. In the affected areas, labor supply decreased
and labor demand increased—producing an increase in relative wages. Our results show substantial differences in

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

the trajectory of earnings by industry: those employed in sectors related to rebuilding had some of the largest
gains, while those employed in local services, education, and healthcare had modest gains (or even losses).
Construction expenditures may have contributed to the recovery of other sectors. By generating demand for local
products and services and providing earnings to local construction workers, construction spending may have
boosted labor demand in other sectors. We find that workers in manufacturing; local services; and trade,
transportation, and utilities had medium-term earnings gains. In contrast, we find no such gains for workers in the
leisure and accommodations, healthcare, and professional services sectors. These sectors are more closely tied to
tourism or the size of the local population.

September 2020

An introduction to behavioral economics: using
psychology to explain economic behavior
Behavioral Economics: The Basics. By Philip Corr and
Anke Plagnol. London and New York: Routledge, 2019, 250
pp., $25.95 paperback.
In this book, authors Philip Corr and Anke Plagnol provide
an introduction to behavioral economics, a relatively new
field of study that uses insights from psychology to
understand economic behavior. Books in behavioral
economics are plentiful and growing in number, ranging
from rigorous and highly technical treatments of various
topics in the field to accessible general-audience popular
books telling stories about how people sometimes don’t
behave as predicted by standard economic models.
Behavioral Economics: The Basics falls within these two
extremes: it summarizes the academic literature related to
behavioral economics and provides a large number of
examples drawing from case studies and anecdotal
evidence.
The book is organized in seven chapters. Chapter 1
introduces the basic concepts of behavioral economics and
explains why this new field is important. The chapter
discusses how behavioral economics uses ideas from
psychology to study economic behavior and argues that
standard economics often fails to explain how people
behave. Corr and Plagnol mention the 2007–08 financial
crisis, a macroeconomic event (at the level of the economy
as a whole), as an example illustrating the limitations of
standard economics, although the book focuses on

Ricardo A. Lopez Rago
ricardo.a.lopezrago@irs.gov
Ricardo A. Lopez Rago, Ph.D., is an economist in
the Large Business and International Division of
the Internal Revenue Service, U.S. Department of
the Treasury.

microeconomics (how individuals make decisions). The
authors go on to characterize standard economic theory,
which is grounded in rational choice, as normative
economics (making judgements on economic policy or
economic behavior) and behavioral economics as positive

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

economics (describing and explaining economic behavior without making judgements). While most economists
won’t agree with these characterizations, some behavioral economists, including Corr and Plagnol, have adopted
this controversial interpretation of normative versus positive economics.
Chapter 1 also discusses the role of behavioral economics in the economics discipline. One view, which I believe
is shared by most economists, is that behavioral economics should be used to complement standard economics.
In that sense, behavioral economics can explain anomalous cases of individual economic behavior. Another view,
and one the authors seem to adopt throughout the book, is that standard economic theory is not only incapable of
explaining economic behavior, but also that economists have not taken the criticisms from behavioral economics
seriously. As one example, the authors note that economists rarely take the results of laboratory studies, which
often use small samples of undergraduate students, as reliable or generalizable. While these studies may provide
some interesting findings, their tiny samples (the number of participants is limited by the size of a classroom or a
computer lab) and the complexity of the decisions asked of students (many of whom are not majoring in
economics) render their experimental results biased and unscientific.
Chapters 2 and 3 present a brief introduction to the history of economic thought. Chapter 2 focuses on selected
ideas proposed by classical economists, such as Adam Smith, David Ricardo, and Thomas Malthus, while chapter
3 focuses on modern economic theory, which the authors refer to as neoclassical economics. While these two
chapters are interesting and provide useful context for the average reader, the ideas discussed in them are well
known and readily available in most introductory economics textbooks.
Chapter 4 is the most informative chapter of the book and presents an excellent summary of the field of behavioral
economics. While the authors base their analysis mostly on anecdotal evidence and small-scale case studies, they
identify and discuss possible explanations for why people misbehave or make mistakes. The chapter starts by
explaining the concept of loss aversion, which can be described as a situation in which the pleasure of receiving,
for example, $100 is less than the displeasure of losing $100. The chapter then describes the endowment effect,
which is valuing something we own more than the things we don’t own, and mental accounting, which refers to our
tendency to assign certain amounts of money to specific uses (some money for food, some for college tuition,
some for a vacation, etc.). The chapter also discusses heuristics, which the authors describe as mental shortcuts
that people use to make quick decisions. One example is anchoring, or the overreliance on initial, incomplete
information, such as a readily available number (the anchor), to estimate the value of an item. The chapter also
describes framing effects, which refer to how people change their behavior on the basis of how information is
presented to them. For example, some people may be more likely to buy ground meat if its packaging describes it
as being 95 percent lean as opposed to containing 5 percent fat, even though the product purchased is identical in
both cases.
In chapters 5, 6, and 7, the authors use ideas from psychology to explain human behavior. As such, these chapters
have less to do with behavioral economics, because they try to explain human behavior in areas not always
related to economics. Although many readers are likely to find this discussion interesting, those who are very busy
may want to skim through it and select specific topics they want to explore in more detail.
Chapter 5 argues that people don’t typically maximize their own objectives and that their behavior is highly
influenced by psychological factors. To support this argument, the authors present a mix of theoretical results
obtained from games such as the prisoner’s dilemma and case-study evidence obtained from games such as the
ultimatum game. One could argue, however, that the most likely reason why people do not always maximize their

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

objectives is lack of information rather than psychology. The chapter also summarizes findings from experiments
using monkeys as subjects.
Chapter 6 discusses if nudges can help people make better decisions and whether governments should use them.
The use of nudges is controversial, and the authors do an excellent job of explaining the different sides of the
debate. Finally, chapter 7 reviews how psychology has influenced marketing and advertising. The book does not
contain a concluding chapter summarizing the main arguments or discussing the policy implications of the
research.
While the book is interesting and easy to read, one is left with the impression that its disapproval of standard
economic theory is overdone. More effort could have been devoted to proposing a theoretical framework
supporting the arguments of the book, rather than summarizing a collection of stories, anecdotal evidence, and
case studies. Although the book is an engaging read, the most convincing modern economics books rely on
persuasive empirical evidence, such as that obtained from large datasets and econometric techniques that can
control for unobserved individual characteristics, and this is an area in which Behavioral Economics: The Basics
slightly disappoints. Still, undergraduate students taking an introductory course in behavioral economics and
casual readers interested in selected microeconomic applications of psychology may find this book interesting.
Disclaimer: The views and opinions presented in this book review are those of the author and do not necessarily
represent those of the Internal Revenue Service or the U.S. government.

September 2020

Assessing multidimensional worker skill levels
and their allocation in the U.S. labor market
Lawrence H. Leith
Economists traditionally viewed inequality in wages and employment primarily as a function of human capital. The
traditional view held that a worker’s education and skill level were the main factors in determining the kind of job
that a worker could obtain. That view has slowly given way to one in which most economists now view workers as
having multiple skills, with the labor market largely determining how those skills are allocated into jobs that require
different kinds and combinations of skills. The broader view of labor markets has led to improved understanding of
wage and employment inequality. Nevertheless, one limitation of the newer view is that economists still tend to
model workers’ skills in a one-dimensional way. In other words, their models include the assumption that each
worker has one generalized skill (or set of skills) and that different jobs vary in their need for that skill. A second
limitation of the newer view is that economists generally overlook or understate the way that most workers improve
the skills they use frequently and lose the ones they seldom use. The standard scalar measures of human capital
that most economists use do not account for these kinds of changes.
In their article, “Multidimensional skills, sorting, and human capital accumulation” (American Economic Review,
August 2020), economists Jeremy Lise and Fabien Postel-Vinay try to help us better understand the matching
process between workers’ various skills and the demand for those skills in the U.S. labor market. The authors
extend a well-known and widely used “search-theoretic model” that economists have traditionally applied under the
one-dimensional view of worker skills and apply it to multidimensional skills. They use occupation-level measures
of skill requirements from the U.S. Department of Labor Occupational Information Network and combine them with
data from the National Longitudinal Survey of Youth 1979. Lise and Postel-Vinay examine three categories of skills
—manual, cognitive, and interpersonal—and use their model to assess the economic costs of “mismatch” in the
labor market.
In their model, the authors attempt to account for the fact that manual, cognitive, and interpersonal skills are very
different kinds of productive qualities. Manual skills generally yield only modest returns and can be adjusted fairly
quickly. Workers accumulate such skills rapidly on the job and lose them when they are not used. Cognitive skills
yield much higher returns than manual skills, but they also take more time to adjust or alter than manual skills.
Most workers acquire their cognitive skills through education, either before entering the workplace or by continuing
their education after acquiring a particular job, especially one that highly values such skills. Interpersonal skills,
which are sometimes referred to as noncognitive skills or even personality traits, are more difficult to measure.
They yield modest returns—more than manual skills but less than cognitive skills—and are essentially fixed over a
worker’s lifetime. Although people can improve their interpersonal skills, for most workers, such skills tend to be
relatively stable and unchanging over the course of their working lives.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Lise and Postel-Vinay find that the costs of mismatching are greatest for cognitive skills. In fact, they estimate that
the costs are higher than those of mismatching for manual or interpersonal skills by an order of magnitude.
Moreover, they find that such costs are unequal, in the sense that employing a worker who is underqualified in
cognitive skills is more than twice as costly, in terms of lost surplus, as employing one who is overqualified. The
authors point out that such subtleties and differences tend to be missed in models that assign a single scalar value
to a worker’s skill level. They compare their use of a multidimensional model with the more common onedimensional model and show that the latter considerably overstates the importance of “unobserved
heterogeneity” (diverse skills the worker has that are not immediately apparent) and understates the importance of
“career shocks” (unexpected life-changing events that are out of the worker’s control).

September 2020

Projections overview and highlights, 2019–29
Employment and real output growth are projected to slow
from 2019 to 2029. One in four people will be ages 65 and
older in 2029, contributing to slower projected growth in the
labor force and a continued decline in the labor force
participation rate. The aging population is expected to
continue to drive strong demand for a variety of healthcare
services, with 3.1 million jobs projected to be added in the
healthcare and social assistance sector through 2029.
The U.S. Bureau of Labor Statistics (BLS) projects 0.4percent annual growth in employment over the 2019–29
decade.[1] This projected growth is slower than the growth
that occurred over the 2009–19 decade, which was marked by faster recovery growth following the trough of the
2007–09 Great Recession. The total economy will add about 6.0 million jobs, with employment reaching a level of
168.8 million in 2029. Various demographic trends, including an aging population, are expected to drive slow
growth in the labor force and a lower labor force participation rate over the projections period. These demographic
trends, combined with slower growth in the civilian noninstitutional population, will affect population and labor
force, aggregate demand, industry output and employment, and occupational employment projections over the
2019–29 decade.
This article presents an overview of the 2019–29 projections. Highlights include the following:
• Labor force growth is projected to be slower (0.5-percent annual growth), in part, from an aging population
and slower population growth among Hispanics.
• The labor force participation rate is projected to continue to decline from 63.1 percent in 2019 to 61.2
percent in 2029.
• Gross domestic product (GDP) is projected to continue to grow (1.8 percent annually), but at a slower rate
than the historical pattern.
• Most employment gains over the 2019–29 period are expected to be in the service-providing sectors, led by
strong growth in the healthcare and social assistance sector. The aging population will continue to create
strong demand for industries and occupations that provide healthcare and related services.
Compared with the prior decade, population growth is expected to slow from 2019 to 2029, in part, because of the
slowed growth among the Hispanic population. The median age of the population will continue to rise, with all baby
boomers reaching ages 65 and older by 2029. This increase in median age and an increase in the number of

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

younger people choosing to pursue education before
entering the labor force are expected to contribute to a

Kevin S. Dubina
dubina.kevin@bls.gov

decline in the labor force participation rate in 2029.
Real output is projected to increase by more than $6.8

Kevin S. Dubina is an economist in the Office of
Occupational Statistics and Employment
Projections, U.S. Bureau of Labor Statistics.

trillion from 2019 to 2029, with most growth expected to
occur in the service-providing sectors. The 1.8-percent
annual growth in output projected for the total economy is
slower than the 2.2-percent annual growth from 2009 to

Janie-Lynn Kim
kim.janie-lynn@bls.gov

2019.
Total employment is projected to grow 0.4-percent annually
from 2019 to 2029, slower than the 1.3-percent annual
growth rate experienced from 2009 to 2019, following the
trough of the 2007–09 recession.[2] By comparison, the

Janie-Lynn Kim is an economist in the Office of
Occupational Statistics and Employment
Projections, U.S. Bureau of Labor Statistics.
Emily Rolen
rolen.emily@bls.gov

average of the 10-year growth rates for each year over the
period 2007 through 2019 was 0.5 percent. Serviceproviding sectors will account for most of the jobs added by

Emily Rolen is an economist in the Office of
Occupational Statistics and Employment
Projections, U.S. Bureau of Labor Statistics.

2029. Of the 6.0 million jobs projected to be added to the
economy, about half (3.1 million) are expected to be in the
healthcare and social assistance sector. Employment
increases in this sector will stem from greater demand for a
variety of healthcare services as the population continues
to age and rates of chronic disease continue to increase.

Michael J. Rieley
rieley.michael@bls.gov
Michael J. Rieley is an economist in the Office of
Occupational Statistics and Employment
Projections, U.S. Bureau of Labor Statistics.

Employment declines are expected in the goods-producing
sectors, with the manufacturing sector leading the losses.
Increasing automation, combined with international competition, will lead to employment declines in the
manufacturing sector and in many of the production occupations concentrated in this sector. Changing consumer
preferences and the increase in the use of technology will lead to declines in employment in the postal service,
retail trade, agriculture, and several information-related industries.

Effects of the COVID-19 pandemic on the 2019–29 projections
The 2019–29 projections do not include impacts of the coronavirus disease 2019 (COVID-19) pandemic and
response efforts. BLS develops the BLS employment projections by using models based on historical data,
which, in this set of projections, cover the period through 2019; therefore, all input data precede the
pandemic. In addition, the 2019–29 projections were finalized in spring 2020 when substantial uncertainty
about the duration and impacts of the pandemic still existed.
The BLS employment projections are long-term projections intended to capture structural change in the
economy, not cyclical fluctuations. As such, they are not intended to capture the impacts of the recession
that began in February 2020. However, besides the immediate recessionary impacts, the pandemic may
cause new structural changes to the economy. BLS releases new employment projections annually, and

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

subsequent projections will incorporate new information on economic structural changes as it becomes
available.
To provide more information about potential impacts before the release of the 2020–30 projections, BLS is
developing alternate scenarios for the 2019–29 projections period that encompass possible impacts from
the pandemic. Comparing these alternate scenarios with the baseline 2019–29 projections released here
will demonstrate how changes in consumer behavior caused by the pandemic may alter the projections for
detailed occupations and industries. An analysis of these scenarios will be released in a Monthly Labor
Review article later in 2020.

Preparing the projections—methodology overview
BLS prepares projections in four areas: population and labor force, aggregate demand, industry output and
employment, and occupational employment. Each step in the projections process affects those that follow. The
expectations for the population affect those for the labor force, which in turn affect the projections for productivity
and GDP growth. These projections further affect output and employment at the industry level, which then limit
occupational employment projections.
BLS makes labor force projections by incorporating U.S. Census Bureau population projections in BLS projections
of the labor force participation rate. In the BLS labor force model, population growth and changes in participation
rates are the main factors in labor force growth. However, most of the changes in labor force growth are due
to changes in the population. The current BLS labor force projections to 2029 are based on the 2018 national
population projections made by the U.S. Census Bureau. The projections include assumptions about future fertility
and mortality rates of the U.S. population. Also included are assumptions about immigration, an important but
uncertain factor affecting the size of the future labor force (immigration data are from the Census Bureau).
Because labor force growth is one of the major determinants of long-term economic growth, labor force projections
describe the future path of the economy and its capacity to create goods and services. The long-term gradual
slowdown in the labor force growth continues to be key in determining the growth of the economy and of
employment.
BLS develops macroeconomic projections with a model licensed from Macroeconomic Advisers (MA) by IHS
Markit.[3] The MA model includes an assumption of full employment in the target year, allowing the projections to
underscore structural changes in the economy rather than cyclical movements in the business cycle. Energy prices
come from the Energy Information Administration (EIA), and BLS determines other critical variables and supplies
them to the MA model exogenously.[4] The MA model then projects economic aggregates, including total
employment, output, productivity, prices, interest rates, and many other variables for the U.S. economy. These
variables, most importantly nonfarm payroll employment, labor productivity, and GDP, serve as constraints for the
industry output and employment projections. These critical variables set the parameters for the nation’s economic
growth and set the trend that GDP will follow and the number of jobs needed to support that trend.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

BLS produces model-based projections for hundreds of detailed industries that are then summed to subsectors
and sectors. Macroeconomic factors, such as the labor force, GDP and its components, and labor productivity,
affect the growth in total employment. Along with projections models for the individual industries, these factors
determine the final projections of industry employment and output.
BLS produces occupational employment projections by analyzing current and projected future staffing patterns
(the distribution of occupations within an industry) in an industry–occupation matrix. Changes in the staffing pattern
for each industry are projected and applied to the final industry projections, yielding detailed occupational
projections by industry. This projected employment matrix includes estimates for 790 occupations across 295
industries.[5]
The current projections, for 2019–29, are the first set released annually. Prior projections were released every 2
years.

Population and labor force
The population and labor force (the number of people ages 16 and older who are either working or actively looking
for work) have been steadily growing over the span of U.S. history.[6] However, the growth for both has slowed
over the last few decades and is expected to continue slowing over the 2019–29 decade. The labor force
participation rate—the percentage of the civilian noninstitutional population in the labor force—has declined since
the start of the 21st century because of the aging baby boom generation and other demographic trends. Slower
growth of the population and the labor force and a continued decline in the participation rate are projected over the
next decade.

Population
Population growth has slowed substantially since the late 1970s, and this trend is projected to continue. (See
figure 1.) The high growth rates in the late 1970s correspond with the start of the “echo boom,” when children of
the baby boomers entered their prime childbearing years.[7] The smaller uptick in the 2000s can be partially
attributed to increased immigration. Between 1999 and 2009, 3.8 percentage points of resident population growth
were attributable to immigration. Immigration then slowed a bit during and after the 2007–09 recession,
contributing only 3.0 percentage points between 2009 and 2019.[8]

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Over the 2000–10 decade, approximately half of all immigrants reported Hispanic ethnicity.[9] The Hispanic
population grew 4.3 percent compounded annually between 1999 and 2009, faster than the 0.9-percent
compounded annual growth in the non-Hispanic population. (See figure 2.) The high Hispanic growth rate was not
strictly because of immigration. The birth rate amongst Hispanic individuals was higher than that of nonHispanics.[10]

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Between 2009 and 2019, the growth among foreign-born individuals living in the United States slowed.[11]
Although the Hispanic population growth rate fell to 2.8 percent compounded annually, it was still significantly
higher than that of the non-Hispanic population. In spite of the immigration slowdown, the Hispanic growth rate
was still responsible for boosting the overall population growth rate by 0.4 percentage point over the same 2009–
19 period.
Slowing immigration is expected to continue to affect overall population growth over the projections period. The
overall population growth rate is projected to decline slightly to 0.8 percent compounded annually from 2019–29.
The slowdown is because of continued decreases in both the rate of Hispanic growth and non-Hispanic growth,
which are expected to fall to 2.4 percent and 0.4 percent compounded annually, respectively. (See figure 2.)
Other demographic shifts are also affecting the U.S. population. A growing percentage of the population is found in
the higher age categories. Baby boomers began entering the 55-and-older cohort in 2001 and the 65-and-older
cohort in 2011. These trends are particularly important for the labor force because older people are less likely to
work than those ages 25 to 54. Through 2029, the growth of those ages 55 and older is expected to slow because
all baby boomers are already in this group. However, the 65-and-older and 75-and-older groups are expected to
continue to see their increased growth rates maintained. (See figure 3.) The growth among those ages 55 and
older has contributed to a lower labor force participation rate and will continue to do so in the future.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Labor force and participation rate
Growth of the labor force has slowed, in large part, because of the two previously discussed trends—the aging
population and slower population growth. A large subset of the population is in the labor force; therefore, the labor
force often takes a cue from population growth and behaves similarly. (See figure 4.)

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

The percentage of the civilian noninstitutional population in the labor force is known as the labor force participation
rate. The steady increase in the participation rate over the latter half of the 20th century was largely driven by
women entering the labor force. Between 1997 and 2000, the overall participation rate peaked at 67.1 percent and
has declined over most of the past two decades.
The participation rate fell steeply, well below potential growth,[12] between 2008 and 2015 in the aftermath of the
2007–09 recession. Since 2016, the participation rate has been edging up closer to, and possibly even surpassing,
its 2019 potential, as estimated by the Congressional Budget Office.[13] Therefore, this uptick in the labor force
participation rate between 2015 and 2019 appears to be due to cyclicality and not a sustainable trend. (See figure
5.) The aging population is expected to be the largest driver of the projected decreasing participation rate because
older individuals are less likely to be in the labor force.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

However, the aging population is not the only reason the overall participation rate is decreasing. The prime-age
labor force participation rate for men ages 25 to 54 has steadily fallen, from 96.1 percent in 1969 to 89.1 percent in
2019, and is projected to continue to decline to 87.3 percent in 2029. (See figure 6.)

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Historically, the participation rate fell the most for men with only a high school degree, some college, or an
associate’s degree and for men on the younger end of the prime-age range (ages 25 to 34). This group is most
likely to be employed in middle-skill jobs, often considered “routine” occupations that have become automated by
computers and machines. In addition, international trade and weakening unions have contributed to a decline in
these middle-skill jobs.[14]
The participation rate of youths, ages 16 to 24, for both genders also has declined. The 16- to 24-year-old
participation rate fell from 65.5 percent in 1999 to 55.9 percent in 2019. It is projected to decline still further, to 51.3
percent in 2029. (See figure 7.) The decline in labor force participation of youth corresponds to a higher fraction of
young people attending school.[15]

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

While prime-age (25 to 54) individuals—particularly men—and youths are working less, older individuals are
working more. The participation rate for the 65-and-older group has been rising since the 1990s and is projected to
continue to rise. (See figure 8.) Several factors are driving this trend, including longer, healthier lifespans and shifts
in retirement programs, which include changes to Social Security.[16] Despite these shifts, people ages 25 to 54
are projected to continue to have much higher participation rates than those ages 65 and older (81.8 percent and
23.4 percent, respectively).

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Macroeconomic projections
The U.S. economic output—which is associated with GDP—has grown throughout history. BLS projects output will
continue growing; however, the 1.8-percent annual compounded growth through 2029 is slower than the rate seen
in recent decades. (See figure 9.) The slower growth in the labor force will result in this slower GDP growth.
However, the labor force is not the only contributor. BLS uses a potential output assumption in the target year
(2029) to remove cyclicality from projections. However, cyclicality can remain in the base year. In 2019, the
economy was at its full potential—so little to no cyclicality existed. This situation is unusual; more often than not,
the economy is below its potential, which gives it more room to grow. This constraint on growth is another reason
GDP growth is expected to grow slower than it has historically.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Figure 9 shows GDP growth over 10-year periods. Through the 1980s and 1990s, GDP growth hovered above 3percent compound annual growth. During the 2007–09 recession, GDP growth plummeted, which is reflected in
the slower growth over the 1999–2009 decade. The 2.3-percent compound annual growth from 2009–19 is not
expected to be a new structural trend. The base year used for calculating this growth (2009) was one of the worst
economic times in this country’s history. In addition, much of the recent 2018 and 2019 increase can be attributed
to the 2017 Tax Cuts and Jobs Act, which is expected to result in a short-term boost, with little further effect on
GDP growth in 2020 or beyond.[17]
Personal consumption expenditures are expected to be the primary driver of GDP growth over the next decade.
This scenario is typical—consumption generally drives the majority of U.S. GDP growth. The exceptions are during
a recession when government spending picks up slack from other sectors and sometimes after a recession when
investment increases to make up for deferred investment during a recession. Over the next 10 years, 1.3
percentage points of annual compound growth are projected to be attributed to growth in consumption and 0.5
percentage point to investment. (See figure 10.)

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Employment, nonaccelerating inflation rate of unemployment, and productivity
The labor force includes not only the employed but also the unemployed. At full employment, the rate of
unemployment is relatively low. However, the unemployment rate never reaches zero—frictional unemployment
will always exist as workers transfer between jobs and seek new jobs. The unemployment rate when the economy
is at full employment is called nonaccelerating inflation rate of unemployment (NAIRU). The unemployment rate, at
NAIRU, is set at 4.4 percent in 2029.[18]
NAIRU and the labor force have important implications for the projection of GDP. Labor productivity also affects
GDP. Labor productivity is measured as the total output divided by hours worked in the economy. Productivity
growth decreased in the wake of the 2007–09 recession. BLS projects productivity to return to a more normal
pattern of growth over the next decade, 1.8-percent compound annual growth, compared with 1.1-percent
compound annual growth in 2009–19. This percent growth is still significantly slower than the 2.7-percent
compound annual growth experienced from 1999 to 2009.
Capital deepening, an increase in the capital to labor ratio, is the largest driver of productivity. Other drivers are
lumped together and are called total factor productivity. These other drivers include technological advances,
education or quality of the workforce, improvements in management practices, and economies of scale. Over the
upcoming decade, capital deepening is projected to make up 1.2 percentage points of productivity compounded
annual growth, compared with 0.9 percentage point for total factor productivity. (See figure 11.)

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Fiscal and monetary policy
The monetary policy goal of the Federal Reserve is to foster economic conditions that achieve both stable prices
and maximum sustainable employment.[19] The Federal Reserve targets 2-percent inflation to achieve its stable
prices mandate.[20] In some environments, balancing these mandates may be challenging. Up to February 2020,
however, the economy has managed to achieve full employment while inflation has been consistent, around 2.0
percent.[21]
The Federal Reserve’s primary tool for managing this goal is through the federal funds rate. The federal funds rate
is the rate at which banks lend money to each other. Consumer borrowing rates are higher than this rate, but they
tend to move with it. The federal funds rate has been trending upward since it was lowered to 0.0 in the wake of
the 2007–09 recession. In 2019, the federal funds rate was 2.2 percent, and it is projected to increase to 2.4
percent in 2029. (See figure 12.) This percentage is significantly below the federal funds rate for much of history—
it reached 5.0 percent before the 2007–09 recession (in 2007) and above 6.0 percent before the 2001 recession
(in 2000).

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

After the 2007–09 recession, there was some concern that low interest rates and an expanding money supply
could cause excess inflation.[22] However, this worry turned out to be unfounded. Over time, the natural rate of
interest can vary in response to shifts in preferences and technology. Evidence shows that such a shift to a lower
natural rate may be occurring.[23] One reason that may explain the change is the aging of the population. As
people live longer, they prefer to save more money to supplant a longer retirement period, which increases the
supply of borrowable money and drives down interest rates.[24]
Assumptions about fiscal policy, including tax policies and government spending, substantially affect expectations
for government revenue, national debt, and economic growth. BLS generally assumes no major changes to current
tax laws over the projections decade. Effective marginal tax rates also are held constant at their current levels.

Energy prices
Energy prices, particularly oil prices, are another macroeconomic variable of interest because of their influence on
consumer spending. Lower prices are generally associated with more economic growth because they free up
additional money for consumer discretionary spending. However, if prices fall exceptionally fast, the energy sector
may contract, which negatively influences GDP. Energy prices used by BLS within the U.S. Macro Model (MA/
U.S.) from IHS Markit are supplied from the EIA’s January 2020 Annual Energy Outlook.[25] From 2019 to 2029,
oil prices are projected to rise approximately 50 percent, from $57 to $89 for West Texas Intermediate (WTI) and
$64 to $95 for Brent. (See figure 13.)

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Industry output and employment projections to 2029
BLS projects real gross industry output will increase slightly more in the 2019–29 decade than it did during the
previous decade, whereas employment growth will be slower. Industry output and employment projections were
prepared with the use of the North American Industry Classification System (NAICS). Major sectors—hereafter
referred to as “sectors”—are aggregations of NAICS industries.

Industry output
BLS projects that real gross industry output will increase from just over $34.0 trillion in 2019 to roughly $40.9 trillion
in 2029.[26] The more than $6.8 trillion increase by 2029 is slightly larger than the increase during the previous
decade. However, overall growth is projected to slow to a 1.8-percent rate over the 2019–29 period from a 2.2percent rate over the 2009–19 recovery period. Most of the increase in real output (74.2 percent) is projected to
come from service-providing sectors.

Sector output
In line with projections for the 2018–28 period, output for service-providing sectors is projected to grow at an
average rate of 2.0 percent per year from 2019 to 2029. (See table 1.) This rate is slower than the 2.3-percent
growth experienced from 2009 to 2019. Over the 2019–29 decade, the projected 2.0-percent growth in output for
service-providing sectors is slightly faster than the 1.8-percent projected growth for the entire U.S. economy. All
service-providing sectors are projected to experience real-output growth over the 2019–29 projections period,
except for the federal government sector, which is projected to decline slightly at a rate of 0.1 percent annually.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

The healthcare and social assistance sector is projected to have the fastest growth rate among service-providing
sectors in 2019–29, with a projected annual growth rate at 2.9 percent.
Table 1. Output by major industry sector, 2009–29
Compound
Billions of chained 2012 dollars annual rate

2009

Special Industries (1)
Residual

(2)

Percent distribution

of change

Industry sector

Total
Goods-producing,
excluding agriculture
Mining
Construction
Manufacturing
Service-providing
excluding special
industries
Utilities
Wholesale trade
Retail trade
Transportation
and warehousing
Information
Financial
activities
Professional and
business
services
Educational
services
Health care and
social assistance
Leisure and
hospitality
Other services
Federal
government
State and local
government
Agriculture, forestry,
fishing, and hunting

Billions of dollars

2019

2029

$27,293.3 $34,049.9 $40,867.6

2009– 2019–

2009

2019

2029

2009

2019

2029

2.2

1.8 $25,021.9 $37,733.5 $53,862.7 100.0 100.0 100.0

6,914.1

8,473.7

9,921.6

2.1

1.6

6,034.9

8,637.5

12,465.7

24.1

22.9

23.1

485.8
1,161.5
5,261.6

635.3
1,418.9
6,384.1

835.5
1,616.3
7,434.0

2.7
2.0
2.0

2.8
1.3
1.5

404.0
1,099.0
4,532.0

549.2
1,728.5
6,359.8

1,037.2
2,284.6
9,143.9

1.6
4.4
18.1

1.5
4.6
16.9

1.9
4.2
17.0

18,646.8

23,486.7

28,546.7

2.3

2.0 17,410.90

26,800.0

38,257.7

69.6

71.0

443.6
1,223.5
1,271.2

458.6
1,885.8
1,842.1

534.5
2,440.7
2,339.3

0.3
4.4
3.8

1.5
2.6
2.4

436.5
1,154.8
1,207.7

505.7
2,051.6
1,936.3

718.6
3,207.1
2,678.5

1.7
4.6
4.8

1.3
5.4
5.1

1.3
6.0
5.0

940.4

1,128.4

1,327.9

1.8

1.6

782.0

1,274.7

1,900.2

3.1

3.4

3.5

1,226.6

1,914.2

2,489.5

4.6

2.7

1,219.7

1,899.2

2,602.6

4.9

5.0

4.8

3,613.6

4,227.0

4,953.4

1.6

3,323.4

5,402.7

7,142.7

13.3

14.3

13.3

2,750.9

3,763.7

4,720.1

3.2

2.3

2,610.2

4,266.1

6,039.6

10.4

11.3

11.2

314.0

313.1

371.3

0.0

1.7

273.5

381.4

586.3

1.1

1.0

1.1

1,829.6

2,402.3 3,201.2

2.8

2.9

1,724.1

2,641.2

4,271.0

6.9

7.0

7.9

1,013.8

1,347.3

1,662.8

2.9

2.1

955.2

1,581.3

2,491.5

3.8

4.2

4.6

563.9

651.6

773.6

1.5

1.7

526.5

765.4

1,111.0

2.1

2.0

2.1

1,125.7

1,118.8

1,112.4

–0.1

1,056.2

1,259.9

1,507.3

4.2

3.3

2.8

2,333.5

2,475.1

2,754.4

0.6

1.1

2,141.3

2,834.4

4,001.4

8.6

7.5

7.4

459.5

571.6

684.5

2.2

1.8

342.8

460.6

771.7

1.4

1.2

1.4

1,273.7

1,505.5

1,671.1

1.7

1.0

1,233.1

1,835.5

2,367.6

4.9

4.4

–0.7

12.4

43.7

—

Notes:
(1) Consists of nonproducing accounting categories to reconcile the input–output system with National Income and Product Accounts.
(2) Residual is shown for the first level only. Subcategories do not necessarily add to higher categories as a by-product of chain-weighting.

Source: U.S. Bureau of Labor Statistics.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Real output in the goods-producing sectors (excluding agriculture) is projected to grow at a rate of 1.6 percent per
year from 2019 to 2029, which is slower than the expected growth rate of 1.8 percent for the overall economy. The
1.6-percent growth rate for the goods-producing sectors is slower than the 2.1-percent increase experienced a
decade earlier, but roughly in line with the overall slowing of output growth. Within the nonagricultural goodsproducing sectors, mining has the fastest projected growth rate—2.8 percent annually for the next 10 years.
The agricultural sector (including forestry, fishing, and hunting) is projected to grow at a rate of 1.8 percent per
year for the 2019–29 projections period, the same as the projected rate for overall output growth. This rate is
slower than the rate of 2.2-percent annual growth experienced a decade earlier by both the agricultural sector and
total output.

Fastest growing output
Within the service-providing sector, the information sector is projected to have 3 of the 20 fastest growing realoutput industries from 2019 to 2029: software publishers; other information services; and data processing, hosting,
and related services. The software publishers industry, in particular, continues to be the fastest growing real-output
industry in the U.S. economy, as more consumers demand software to accommodate lifestyle needs, such as
Internet of Things,[27] network integration, cloud computing, and web services. The software publishers industry is
projected to grow at a rate of 4.8 percent annually over the 2019–29 projections period.
The healthcare and social assistance sector includes 10 of the 20 industries with the fastest growing real output for
the 2019–29 projections period. Within healthcare and social assistance, offices of physicians, outpatient care
centers, other ambulatory healthcare services, and medical and diagnostic laboratories industries are projected to
grow the fastest. The aging of the population and the continued expected rise in chronic health conditions, such as
diabetes, are expected to drive demand for healthcare services overall.

Most rapidly declining output
The manufacturing sector includes 3 of the 10 industries projected to decline in real output, with the tobacco
manufacturing industry projected to have the fastest annual rate of decline over the next decade (–2.1 percent).
The continued decline in the number of people who use tobacco products is one of the reasons for the industry’s
drop in real output. The alumina and aluminum production and processing industry and the textile mills and textile
product mills industry, both in the manufacturing sector, are the second- and fourth-fastest declining industries
overall. Increased outsourcing to overseas production for lower labor costs contributes to the overall decline in
manufacturing.
The federal government sector includes 6 of the 10 industries that are projected to decline in real output, in part,
because of continued pressure to reduce government spending.[28] Of all industries, the postal service is
projected to have the third-largest decline in real output, decreasing by 0.8 percent annually over the next decade.
The continued increase in the use of alternative methods of communication—such as email, electronic bill
payment, and digital subscriptions, to name a few—contribute to the decline in real output.

Industry employment projections
BLS uses projected output and labor force data to create projections of total employment by industry. BLS projects
total employment in 2029 to reach 168.8 million, an increase from 2019 of about 6.0 million. This growth
represents a 0.4-percent annual growth rate, which is lower than the 1.3-percent growth rate experienced from the

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

2009 recession trough to 2019. Most of the increase in employment stems from nonagricultural wage and salary
workers: the number of nonagricultural wage and salary jobs is projected to rise from 151.7 million in 2019 to 158.1
million in 2029, an increase of about 6.4 million jobs.[29] (See figure 14.) This increase is less than the 19.7 million
jobs that were added from 2009 to 2019. The 2019–29 employment increase for nonagricultural wage and salary
workers, at a growth rate of 0.4 percent per year, is projected to be slower than the 1.4-percent annual growth rate
experienced from 2009 to 2019.

Sector employment
Employment in service-providing sectors is projected to increase by roughly 6.5 million jobs, reaching about 137.2
million by 2029. This projected increase represents most of the jobs to be added over the 2019–29 projections
period. Employment in service-providing sectors is expected to grow by 0.5 percent annually over the next decade,
slightly faster than the 0.4-percent annual growth for the overall economy. (See table 2.) This growth, however, is
slower than the 1.4-percent annual growth rate experienced by the service-providing sectors from 2009 to 2019.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Table 2. Employment by major industry sector, 2009–29
Compound
Thousands of jobs

Change

Percent distribution annual rate
of change

Industry sector
2009
Total (1)
Nonagriculture wage and salary
Goods-producing, excluding
agriculture
Mining
Construction
Manufacturing
Services-providing excluding
special industries
Utilities
Wholesale trade
Retail trade
Transportation and
warehousing
Information
Financial activities
Professional and business
services
Educational services
Health care and social
assistance
Leisure and hospitality
Other services
Federal government
State and local government
Agriculture, forestry, fishing, and
hunting (3)
Agriculture wage and salary
Agriculture self-employed
Nonagriculture self-employed

(2)

2019

2029

2009–19 2019–29 2009

2019

2029

2009– 2019–
19

143,036.4 162,795.6 168,834.7 19,759.2 6,039.1 100.0 100.0 100.0

1.3

0.4

132,029.2 151,709.7 158,115.6 19,680.5 6,405.9

92.3

93.2

93.7

1.4

0.4

18,507.7

21,016.3

20,964.9

2,508.6

–51.4

12.9

12.4

1.3

0.0

643.3
6,016.5
11,847.9

684.6
7,492.2
12,839.5

777.8
7,792.4
12,394.7

41.3
1,475.7
991.6

93.2
300.2
–444.8

0.4
4.2
8.3

0.4
4.6
7.9

0.5
4.6
7.3

0.6
2.2
0.8

1.3
0.4
–0.4

113,521.5 130,693.4 137,150.7 17,171.9 6,457.3

79.4

80.3

81.2

1.4

0.5

560.1
5,520.9
14,527.6

549.0
5,903.4
15,644.2

506.7
5,801.3
15,275.9

–11.1
382.5
1,116.6

–42.3
–102.1
–368.3

0.4
3.9
10.2

0.3
3.6
9.6

0.3
3.4
9.0

–0.2
0.7
0.7

–0.8
–0.2
–0.2

4,224.7

5,618.1

5,944.1

1,393.4

326.0

3.0

3.5

2.9

0.6

2,803.8
7,838.0

2,859.4
8,746.0

2,853.2
8,799.9

55.6
908.0

–6.2
53.9

2.0
5.5

1.8
5.4

1.7
5.2

0.2
1.1

0.0
0.1

16,633.8

21,313.1

22,831.4

4,679.3 1,518.3

11.6

13.1

13.5

2.5

0.7

3,090.5

3,764.5

4,230.0

465.5

2.2

2.3

2.5

2.0

1.2

16,539.8

20,412.6

23,491.7

3,872.8 3,079.1

11.6

12.5

13.9

2.1

1.4

13,077.5
6,150.1
2,832.0
19,722.7

16,575.9
6,713.8
2,834.0
19,759.4

17,691.5
6,994.7
2,650.4
20,080.0

3,498.4 1,115.6
563.7
280.9
2.0 –183.6
36.7
320.6

9.1
4.3
2.0
13.8

10.2
4.1
1.7
12.1

10.5
4.1
1.6
11.9

2.4
0.9
0.0
0.0

0.7
0.4
–0.7
0.2

2,011.9

2,303.6

2,265.1

291.7

–38.6

1.4

1.3

1.4

–0.2

1,175.7
836.2
8,995.3

1,565.2
738.4
8,782.3

1,600.5
664.5
8,454.1

389.5
–97.8
–213.0

35.3
–73.9
–328.2

0.8
0.6
6.3

1.0
0.5
5.4

0.9
0.4
5.0

2.9
–1.2
–0.2

0.2
–1.0
–0.4

674.0

Notes:
(1) Employment data for wage and salary workers are from the BLS Current Employment Statistics survey, which counts jobs, whereas self-employed and

agriculture, forestry, fishing, and hunting are from the Current Population Survey (household survey), which counts workers.
(2) Includes wage and salary data from the Current Employment Statistics survey, except private households, which is from the Current Populations Survey.

Logging workers are excluded.
(3) Includes agriculture, forestry, fishing, and hunting data from the Current Population Survey, except logging, which is from Current Employment Statistics

survey. Government wage and salary workers are excluded.
Source: U.S. Bureau of Labor Statistics.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

In line with the last five sets of projections, the healthcare and social assistance sector is projected to have the
most employment growth of any sector over the next decade. Employment in healthcare and social assistance is
projected to add about 3.1 million jobs over the 2019–29 decade, reaching a level of 23.5 million by 2029. The
sector is projected to grow at an annual rate of 1.4 percent, more than 3 times as fast as the annual growth for the
overall economy. The projected growth for the healthcare and social assistance sector, however, is still slower than
the 2.1-percent annual growth experienced during the 2009–19 period.
Conversely, the retail trade industry is projected to have the largest employment decline among all serviceproviding industries. Employment in the retail trade industry is projected to decline by 368,300 jobs from 2019 to
2029, a sharp contrast from the previous decade when it added 1.1 million jobs. The declining trend in retail trade
employment is driven by several factors, including bankruptcy and consolidation of big box stores and the shift of
consumer-spending behavior in favor of e-commerce shopping.[30]
Overall employment in the goods-producing sectors (excluding agriculture) is projected to decline over the 2019–
29 projections period. These sectors experienced modest gains from 2009 to 2019 (+2.5 million jobs), offsetting
larger losses experienced during the decade prior (1999–2009). Employment in the construction sector is
projected to increase by 300,200 from 2019 to 2029, growing at an annual rate of 0.4 percent. This increase is
much smaller than the job gains experienced during the previous decade, when construction added nearly 1.48
million jobs as the sector recovered following steep losses during the Great Recession. (See figure 15.) The
manufacturing sector, the largest sector within the goods-producing sectors (excluding agriculture), accounts for
over half of total employment in the goods-producing sectors. The manufacturing sector is projected to decline by
444,800 jobs over the next decade, overshadowing increases in the mining and construction industries. During the
previous decade, manufacturing added 991,600 jobs. (See figure 16.)

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

The agriculture, forestry, fishing, and hunting sector is projected to decline by 38,600 jobs from 2019 to 2029.
During the previous 2009–19 decade, this sector added 291,700 jobs. The projected decline is largely due to a
combination of the slowed employment growth in the crop production industry and a continued decline in the
number of self-employed workers within the agriculture, forestry, fishing, and hunting sector. Employment in the
crop production industry is projected to increase by 66,200 jobs for the 2019–29 decade, whereas 336,100 jobs
were added to the industry during the previous decade. Along with employment growth in the crop production
industry that is projected to slow, the number of self-employed workers in the agriculture, forestry, fishing, and
hunting sector is projected to decline by 73,900 over the next decade, further exacerbating the loss of selfemployed jobs that occurred in this industry during the last decade. This loss is due, in part, to the overall declining
number of small farms, to the emergence of large farming operations, and to older workers being more likely to be
self-employed than any other working age group in this industry.[31]

Fastest growing industry employment
Although overall agriculture, forestry, fishing, and hunting sector employment is projected to decline from 2019 to
2029, employment in the forestry industry component is projected to grow the fastest among all industries for the
2019–29 projections period. The forestry industry is projected to grow at an annual rate of 3.7 percent. However,
because of the industry’s small size, its fast projected growth does little to offset the declines expected in
agriculture, forestry, fishing, and hunting sector employment. In the previous decade, the forestry industry declined
at a rate of 2.4 percent annually.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Over the next decade, 5 out of the 20 fastest growing industries are in the healthcare and social assistance sector:
individual and family services, home healthcare services, outpatient care centers, offices of other health
practitioners, and other ambulatory healthcare services. Factors that contribute to the large projected increase in
the number of healthcare and social assistance jobs include increased demand to care for the aging of the baby
boom population, longer life expectancies, and continued growth in the number of patients with chronic
conditions.32 At an annual growth rate of 3.4 percent, employment in the individual and family services industry is
projected to be the second-fastest growing industry overall.
For the 2019–29 projections period, three industries within the professional and business services sector are
projected to be among the fastest growing industries overall. Within this sector, computer systems design and
related services; management, scientific, and technical consulting services; and office administrative services are
projected to experience fast job gains. Increased demand for technology and the growing complexity of business
operations will contribute to the overall fast employment growth in professional and business services.[33]

Most rapidly declining industry employment
The manufacturing sector is projected to lose the most jobs and have the most rapid employment decline of all
sectors over the 2019–29 projections period. The manufacturing sector includes 12 of the 20 industries projected
to have the largest job declines. Increased international competition and continued automation that increases
overall production with fewer workers will continue to contribute to the loss of jobs for this sector.34 The tobacco
manufacturing industry is projected to have the most rapid declines in industry employment, falling by 5.2 percent
annually.
The information sector includes 3 of the 20 overall most rapidly declining industries for the 2019–29 decade. The
first one is the cable and other subscription programming industry, which is the second-fastest declining industry
within this sector for the projections period, and is projected to decline by 4.8 percent annually from 2019 to 2029.
The second one is the newspaper, periodical, book, and directory publishers industry and is projected to decline by
4.0 percent annually. The wired telecommunications carriers, the third industry, is projected to lose jobs at an
annual rate of 2.1 percent for the same period. The expectation of continued technological changes leading to
fewer job opportunities, the continued trajectory toward digital readership versus print subscription, and a decline
in the number of overall subscriptions will contribute to these employment declines.35

Occupational projections of major groups
To construct projections by occupation, BLS combines the projected total industry employment with staffing-pattern
information. BLS uses the Standard Occupational Classification (SOC) system to categorize occupations in 22
major groups.[36] Occupations are classified in the SOC system on the basis of the type of work performed, job
tasks, and job duties. Examples include statisticians, mathematicians, computer programmers, and web
developers, and all are in the broader computer and mathematical occupational group.
Healthcare support is the fastest growing occupational group, with a projected growth rate of 22.6 percent. (See
figure 17.) Increased demand for healthcare and related employment is also reflected in the high projected growth
rates for healthcare practitioners and technical occupations and community and social service occupations.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Healthcare support occupations include home health and personal care aides, the detailed occupation projected to
add the most new jobs from 2019 to 2029.[37]

Other occupational groups in which employment is projected to grow markedly faster than the average for all
occupations (3.7 percent) include computer and mathematical occupations, personal care and service
occupations, and food preparation and serving related occupations. Computer occupations are expected to see job
growth as strong demand is expected for IT (information technology) security and software development and as
new products associated with the Internet of Things are developed. Rising incomes and a higher share of
expenditures on food away from home are expected to drive growth for food preparation workers.[38]
Four major occupational groups are expected to lose employment: office and administrative support occupations,
with a projected decline of 4.7 percent over the decade; production occupations, with a projected decline of 4.5
percent; sales and related occupations, with a projected decline of 2.0 percent; and farming, fishing, and forestry
occupations, with a projected decline of 0.1 percent. Although employment declines among sales and related
occupations will result from increasing competition from e-commerce, declines in the other groups reflect the
increasing adoption of automation and productivity-enhancing technology in clerical and administrative work,
manufacturing, and agriculture.

Fastest growing occupational employment

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

The projected fast employment growth in the healthcare and social assistance sector is expected to increase
employment substantially in many healthcare occupations from 2019 to 2029. (See figure 18.) Healthcare
occupations and those associated with healthcare (including mental health) account for 13 of the 30 fastest
growing occupations from 2019 to 2029. Increased demand for healthcare services by aging baby boomers and
people with chronic conditions will drive the projected employment growth.[39] Several of the fastest growing
healthcare occupations—including nurse practitioners, occupational therapy assistants, and physician assistants—
are projected to be in greater demand because team-based healthcare models are increasingly used to deliver
healthcare services.[40]

Within the community and social service occupational group, two healthcare-related counseling occupations are
also projected to grow rapidly. Substance abuse, behavioral disorder, and mental health counselors and marriage
and family therapists are projected to have fast employment growth because of increased demand for treatment of
mental and behavioral issues, including opioid addiction. Healthcare-associated occupations from the
management and education occupational groups—medical and health services managers and postsecondary
health specialties teachers—are also expected to be among the fastest growing occupations.
Growth in information and related computer industries is expected to drive employment growth for several
occupations in the computer and mathematical group. This group contains 5 of the 30 fastest growing occupations.
As more devices are connected to the internet, the need to combat cybersecurity threats will increase. The risk of
cyberattacks is expected to create demand for information security analysts, who will be needed to prevent the
theft of critical information and to prevent service attacks on computer networks. Employment of these analysts is

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

projected to increase 31.2 percent from 2019 to 2029. The expected increased use of mobile devices and software
to operate or manage everything from home appliances to medical devices will create demand for software
developers and software quality assurance analysts and testers. Employment in this occupation is projected to
grow 21.5 percent over the decade. Increased use of mobile devices will also drive demand for web-based and
application-based video content, which in turn will lead to employment demand for film and video editors. This
occupation’s employment is projected to grow 21.6 percent from 2019 to 2029.
Employment is projected to grow for statisticians, data scientists, and operations research analysts because of the
increasing widespread use of statistical analysis to make informed business, healthcare, and policy decisions.
(See figure 19.) In addition, the growing amount of data available online (“big data”) will open new areas for
analysis for these occupations.

Two of the top three projected fastest growing occupations—wind turbine service technicians and solar
photovoltaic (PV) installers—are involved in alternative energy production. Employment for wind turbine service
technicians is expected to grow extremely fast (60.7 percent) from 2019 to 2029 as the expansion and adoption of
wind turbines and their installation create new jobs. However, because this occupation is relatively small, with a
2019 employment level of 7,000, the fast growth will account for only about 4,300 new jobs over the next 10 years.
In addition, developments in solar energy generation have made solar energy increasingly competitive with
traditional power generation sources, such as coal and natural gas, and are expected to drive employment growth

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

for PV installers. Employment of these workers is projected to grow 50.5 percent over the next 10 years. Like wind
turbine service technicians, this occupation is small and its rapid growth will account for only about 6,100 new jobs.

Most rapidly declining occupational employment
As noted earlier, the manufacturing sector is projected to lose the most jobs and has one of the most rapid
employment declines of any sector over the projections decade. The decline in employment in the manufacturing
sector is expected to decrease employment over the projections decade in several occupations concentrated in
manufacturing. Production occupations are projected to experience the second-strongest employment decline of
any occupational group, because of a combination of productivity-enhancing technologies, such as robotics, and
international competition.[41] Of the 30 occupations with the fastest employment declines, 12 are in the production
occupational group and include various machine and tool setters, assemblers, and operators. Although their
employment is projected to decline rapidly, they are relatively small occupations and are projected to lose 45,800
jobs, in total.
Similarly, technological changes are expected to continue to negatively affect the employment of several office and
administrative support occupations. For example, software tools can help schedule meetings and appointments
(replacing secretaries and administrative assistants), and digital data collection and handwriting recognition
software can perform work previously done by data entry keyers. Of the 30 occupations with the fastest declining
employment, 8 are office and administrative support occupations. Collectively, these 8 occupations are projected
to lose about 257,400 jobs from 2019 to 2029.
Sales and related occupations are also expected to decline in employment over the next decade, largely because
of competition from e-commerce activity and automation of checkout positions. Cashiers, retail salespersons, and
first-line supervisors of retail sales workers are each projected to see employment declines from 2019 to 2029, for
a combined loss of 371,600 jobs, as online shopping displaces brick-and-mortar retail employment.[42]
Farming, fishing, and forestry occupations are also projected to decline, by 0.1 percent from 2019 to 2029.
Consolidation in the agricultural industry—a greater share of farming output moving from small to large farms—is
allowing more agricultural output to be produced with fewer workers. In addition, automating technology, such as
robotics, is reducing employment demand for farm laborers.[43]

Conclusions
An aging population and slower population growth will result in slower growth in the labor force from 2019 to 2029
than in prior decades. Older people participate in the labor force less than younger people do, so fewer people are
available to be employed. As a result, the projected growth for all jobs, at 3.7 percent, is slower than it was during
the prior projections decade. In addition, since the base year of the projections (2019) is after a long economic
expansion, economic growth rates are expected to be lower than rates in previous projection cycles.
From 2019 to 2029, employment in the service-providing sectors is projected to grow while that in the goodsproducing sector is projected to decline. Occupations that provide healthcare or services related to healthcare are
projected to be the most represented among the fastest growing occupations. An aging population is projected to
demand more healthcare and related services. In addition, the number of people with chronic conditions is

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

projected to continue to grow, adding to the demand for healthcare-related occupations. Other occupations
projected to grow rapidly include those involved with computers, math, and alternative energy.
SUGGESTED CITATION

Kevin S. Dubina, Janie-Lynn Kim, Emily Rolen, and Michael J. Rieley, "Projections overview and highlights, 2019–
29," Monthly Labor Review, U.S. Bureau of Labor Statistics, September 2020, https://doi.org/10.21916/mlr.
2020.21.
NOTES
1 Annual growth refers to a compounded annual growth rate.
2 Total employment is the summation of the employment figures for nonagricultural wage and salary workers; agricultural, forestry,
fishing, and hunting workers; and self-employed workers. Nonagricultural wage and salary employment data are from the U.S. Bureau
of Labor Statistics (BLS) Current Employment Statistics (CES) survey, excluding data for logging, and include private household
employment data, which are provided by the Current Population Survey (CPS). The CPS also provides data for self-employed
workers and agricultural, forestry, fishing, and hunting workers, except data for logging workers, which are provided by the CES
survey.
3 BLS develops macroeconomic projections with the Macroeconomic Advisers (MA) model, a structural econometric model of the
U.S. economy. The model, licensed from MA by IHS Markit, comprises more than 1,000 variables, behavioral equations, and
identities. Central characteristics of the MA model are a life-cycle model of consumption, a neoclassical view of investment, and a
vector autoregression for the monetary policy sector of the economy. The full-employment foundation of the model is the most critical
characteristic for the BLS outlook. Within MA, a submodel calculates an estimate of potential output from the nonfarm business
sector. The calculation is based on full-employment estimates of the sector’s hours worked and output per hour. Error-correction
models are embedded in the MA model so that the model’s solution is aligned with the full-employment submodel. MA does not
forecast sharp cyclical movements in the economy over the 10-year projection horizon. “Add-factors” are either left unchanged after
the first couple of years of the solution or returned to historical norms. Add-factors represent changes made to the base result of a
forecast or projection equation; see “Glossary of statistical terms” (Paris: Organisation for Economic Co-operation and Development,
September 25, 2001, updated March 28, 2014), https://stats.oecd.org/glossary/detail.asp?ID=44. The structure of the model,
exogenous assumptions, and MA’s view of the Federal Reserve’s long-term policy objective largely determine the characteristics of
the model’s long-term outlook for the economy. For more information, see http://www.macroadvisers.com/.
4 Energy Information Administration (EIA) estimates include prices for West Texas Intermediate crude oil, Brent crude oil, and natural
gas and assume current energy regulations will remain unchanged. For more information, see Annual energy outlook 2020 (U.S.
Energy Information Administration, January 29, 2020, released annually), https://www.eia.gov/outlooks/aeo/.
5 Visit our “Employment projections methods overview” page for a detailed description of the projections process at https://
www.bls.gov/emp/documentation/projections-methods.htm.
6 For pre-1946 data, see https://www2.census.gov/library/publications/1949/compendia/hist_stats_1789-1945/hist_stats_1789-1945chD.pdf, and for post-1946 data, see https://www.bls.gov/cps/.
7 Population Research Institute staff, “U.S. birth ‘echo boom’ fades away” (Population Research Institute, September 1, 1995), https://
www.pop.org/u-s-birth-echo-boom-fades-away/.
8 “Foreign born CPS data tables,” table 1.1 for 1999, 2009, and 2019 (U.S. Census Bureau), https://www.census.gov/topics/
population/foreign-born/data/tables/cps-tables.html. Note that these U.S. Census Bureau data encompass the resident population.
The percentage points attributed to noninstitutional population growth are higher because few immigrants are children (ages 0–15), of
whom are included in the resident population.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

9 Eileen Patten, “2010, foreign-born population in the United States statistical portrait,” “Table 6. Population, by nativity, race and
ethnicity: 2010,” Hispanic Trends, February 21, 2012, https://www.pewresearch.org/hispanic/2012/02/21/2010-statistical-informationon-immigrants-in-united-states/.
10 “U.S. fertility rates higher among minorities” (Washington, DC: Population Reference Bureau, January 1, 2003), https://
www.prb.org/usfertilityrateshigheramongminorities/.
11 Sandra Leigh Johnson, “Drops in natural increase, net international migration resulted in 0.5% annual growth to 328.2M,”
Population (U.S. Census Bureau, December 30, 2019), https://www.census.gov/library/stories/2019/12/new-estimates-show-uspopulation-growth-continues-to-slow.html.
12 For a more nuanced discussion on potential growth and its reflection in BLS projections, see Kevin S. Dubina, “Full employment:
an assumption within BLS projections,” Monthly Labor Review, November 2017, https://www.bls.gov/opub/mlr/2017/article/fullemployment-an-assumption-within-bls-projections.htm.
13 Joshua Montes, “CBO’s projection of labor force participation rates,” “Figure 10. Aggregate labor force participation rate,” Working
Paper 2018-04 (Congressional Budget Office, March 2018), p. 20, https://www.cbo.gov/system/files/115th-congress-2017-2018/
workingpaper/53616-wp-laborforceparticipation.pdf.
14 Didem Tüzemen, “Why are prime-age men vanishing from the labor force?” Economic Review (Federal Reserve Bank of Kansas
City, first quarter 2018), https://www.kansascityfed.org/~/media/files/publicat/econrev/econrevarchive/2018/1q18tuzemen.pdf.
15 Maria E. Canon, Marianna Kudlyak, and Yang Liu, “Youth labor force participation continues to fall, but it might be for a good
reason,” Regional Economist (Federal Reserve Bank of St. Louis, January 26, 2015), https://www.stlouisfed.org/publications/regionaleconomist/january-2015/youth-labor-force.
16 Martin Neil Baily and Benjamin H. Harris, “Working longer policies: framing the issues,” Economic Studies at Brookings (The
Brookings Institution and Northwestern Kellogg School of Management, January 2019), https://www.brookings.edu/wp-content/
uploads/2019/01/ES_20180124_Harris-Baily-Retirement-Proposals1.pdf.
17 Karel Mertens, “U.S. tax cuts boost economy—but for how long?” Dallas Fed Economics (Federal Reserve Bank of Dallas, June 4,
2019), https://www.dallasfed.org/research/economics/2019/0604.
18 BLS used the nonaccelerating inflation rate of unemployment estimate as published by the Congressional Budget Office at the
time of our estimation.
19 “The Federal Reserve’s dual mandate” (Federal Reserve Bank of Chicago, updated August 10, 2020), https://www.chicagofed.org/
research/dual-mandate/dual-mandate.
20 “Why does the Federal Reserve aim for inflation of 2 percent over the longer run?” FAQs (Board of Governors of the Federal
Reserve System, Federal Reserve, updated August 27, 2020), https://www.federalreserve.gov/faqs/economy_14400.htm.
21 A variety of inflation measures exist. BLS includes two of the more common measures, the Consumer Price Index and a GDP
index (sometimes referred to as a deflator) in table 4.2. See table 4.2 on the employment projections public website at https://
www.bls.gov/emp/tables/real-gdp-major-demand-category.htm.
22 Arthur B. Laffer, “Get ready for inflation and higher interest rates,” The Wall Street Journal, June 11, 2009, https://www.wsj.com/
articles/SB124458888993599879.
23 For more information regarding the natural rate of interest, see Thomas Laubach and John C. Williams, “Measuring the natural
rate of interest” (Board of Governors of the Federal Reserve System, November 2001), https://www.federalreserve.gov/pubs/feds/
2001/200156/200156pap.pdf; and “Measuring the natural rate of interest” (Economic Research, Federal Reserve Bank of New York),
https://www.newyorkfed.org/research/policy/rstar.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

24 Sungki Hong and Hannah G. Shell, “Factors behind the decline in the U.S. natural rate of interest,” Economic Synopses (Economic
Research, Federal Reserve Bank of St. Louis, April 19, 2019) https://research.stlouisfed.org/publications/economic-synopses/
2019/04/19/factors-behind-the-decline-in-the-u-s-natural-rate-of-interest.
25 For more information, see “Annual energy outlook 2020,” https://www.eia.gov/outlooks/aeo/. This model was run in early 2020,
before oil prices fell dramatically in the wake of the coronavirus disease 2019 stay-at-home order.
26 Throughout this article, output refers to real output in chain-weighted 2012 dollars.
27 “Internet of Things,” Merriam-Webster, https://www.merriam-webster.com/dictionary/Internet%20of%20Things.
28 “Employment projections,” “Table 2.5 Industries with the fastest growing and most rapidly declining output” (U.S. Bureau of Labor
Statistics, last modified September 1, 2020), https://www.bls.gov/emp/tables/industries-fast-grow-decline-output.htm.
29 Nonagricultural wage and salary employment data are from the CES survey, except for private household employment data, which
are from the CPS. Logging workers are excluded.
30 Julia La Roche, “Here’s why the retail sector keeps bleeding jobs,” Yahoo Finance, October 4, 2019, https://finance.yahoo.com/
news/retail-loses-11000-jobs-in-september-145131614.html; Nathaniel Popper, “Americans keep clicking to buy, minting new online
shopping winners,” The New York Times, May 13, 2020, https://www.nytimes.com/interactive/2020/05/13/technology/online-shoppingbuying-sales-coronavirus.html; and Austan Goolsbee, “Never mind the internet. Here’s what’s killing malls,” The New York Times,
February 13, 2020, updated February 14, 2020, https://www.nytimes.com/2020/02/13/business/not-internet-really-killing-malls.html.
31 Steven F. Hipple and Laurel A. Hammond, “Self-Employment in the United States,” Spotlight on Statistics, March 2016, https://
www.bls.gov/spotlight/2016/self-employment-in-the-united-states/pdf/self-employment-in-the-united-states.pdf.
32 Emily Richards Rolen, “Healthcare jobs you can get without a bachelor’s degree,” Beyond the Numbers, November 2016, https://
www.bls.gov/opub/btn/volume-5/pdf/healthcare-jobs-you-can-get-no-bachelors-degree.pdf; and Lauren Medina, Shannon Sabo, and
Jonathan Vespa, “Living longer: historical and projected life expectancy in the United States, 1960 to 2060” (U.S. Census Bureau,
February 2020), https://www.census.gov/content/dam/Census/library/publications/2020/demo/p25-1145.pdf.
33 Zach Lazzari, “What are the causes of rapid growth in the service industry?” Chron, January 22, 2019, https://
smallbusiness.chron.com/causes-rapid-growth-service-industry-16007.html.
34 Alexa Lardieri, “Robots will replace 20 million jobs by 2030, Oxford report finds,” U.S. News and World Reports, June 26, 2019,
https://www.usnews.com/news/economy/articles/2019-06-26/report-robots-will-replace-20-million-manufacturing-jobs-by-2030.
35 Joseph Ahrens, “The decline in newspapers: a closer look,” Wake Review (Raleigh, NC: Wake Review Literary Magazine & Club,
Wake Tech Community College, November 2016), https://clubs.waketech.edu/wake-review/magazine/creative-writing/non-fiction/thedecline-in-newspapers-a-closer-look-joseph-ahrens/; and Michael Barthel, “Newspapers fact sheet,” Journalism & Media (Pew
Research Center, July 9, 2019), http://www.journalism.org/fact-sheet/newspapers/.
36 The 2019–29 National Employment Matrix is based off the hybrid structure between the 2010 and 2018 Standard Occupational
Classification (SOC) systems developed for the Occupational Employment Statistics survey. For more information on the
implementation of the 2018 SOC system, see https://www.bls.gov/oes/soc_2018.htm.
37 The occupation of personal care aides was formerly a separate occupation (code 39-9021) in the 2010 SOC system. In the 2018
SOC system, these workers share a combined code with home health aides and move from the “personal care and service”
occupational group to the “healthcare support” occupational group.
38 Allie Van Duyne, “More people are choosing to dine out. Here’s why.” Toast, https://pos.toasttab.com/blog/why-more-people-arechoosing-to-eat-out.
39 For more information, see Rolen, “Healthcare jobs you can get without a bachelor’s degree.”

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

40 “Creating patient-centered team-based primary care,” Publication no. 16-0002-EF (Agency for Healthcare Research and Quality,
U.S. Department of Health and Human Services, March 2016), https://pcmh.ahrq.gov/page/creating-patient-centered-team-basedprimary-care.
41 Daron Acemoglu and Pascual Restrepo, “Robots and jobs: evidence from US labor markets,” Working Paper 23285 (Cambridge,
MA: National Bureau of Economic Research, March 2017), http://www.nber.org/papers/w23285.pdf.
42 Wolf Richter, “Here’s which brick-and-mortar retailers are getting hit the hardest,” Business Insider, May 19, 2018, https://
www.businessinsider.com/brick-and-mortar-retailers-getting-hit-the-hardest-2018-5.
43 Knvul Sheikh, “A growing presence on the farm: robots,” The New York Times, February 13, 2020, https://www.nytimes.com/
2020/02/13/science/farm-agriculture-robots.html.

RELATED CONTENT

Related Articles
Projections overview and highlights, 2018–28, Monthly Labor Review, October 2019.
Full employment: an assumption within BLS projections, Monthly Labor Review, November 2017.
Projections overview and highlights, 2016–26, Monthly Labor Review, October 2017.
Labor force projections to 2024: the labor force is growing, but slowly, Monthly Labor Review, December 2015.
The U.S. economy to 2024, Monthly Labor Review, December 2015.
Occupational employment projections to 2024, Monthly Labor Review, December 2015.

Related Subjects
Projections

Jobseeking methods

development and growth

Labor dynamics

Labor and economic history

Labor force
Labor market

Unemployment
Recession

Employment
Population

Economic

September 2020

A new BLS satellite series of net inputs to industry
price indexes: methodology and uses
This article describes U.S. Bureau of Labor Statistics efforts
to develop a new set of satellite net inputs to industry price
indexes that capture price change for both domestically
produced and imported inputs to production. In addition to
detailing the methodology for constructing the new indexes,
the article discusses their publication structure and potential
uses.
In September 2020, the U.S. Bureau of Labor Statistics
(BLS) introduced a new set of satellite net inputs to industry
price indexes.1 These indexes measure price changes for
both domestically produced and imported inputs (excluding
capital investment and labor) consumed by most three-digit
North American Industry Classification System (NAICS)
industry groups.2 The new indexes are calculated by using
the detailed price indexes published with the BLS Principal
Federal Economic Indicators of the Producer Price Index
(PPI) program and the International Price Program (which
produces the U.S. Import and Export Price Indexes).

Jayson Pollock
pollock.jayson@bls.gov
Jayson Pollock is a supervisory economist in the
Office of Prices and Living Conditions, U.S.
Bureau of Labor Statistics.
Jonathan C. Weinhagen
weinhagen.jonathan@bls.gov

The new satellite series, published on the PPI webpage, is
the culmination of a long-term BLS effort to develop a
comprehensive set of net inputs to industry price indexes.
BLS first proposed calculating such indexes in the late

Jonathan C. Weinhagen is an economist in the
Office of Prices and Living Conditions, U.S.
Bureau of Labor Statistics.

1970s and published its first series in 1986. The scope of
these indexes and the methodology for calculating them
were relatively limited, because the indexes were only
published for the construction sector of the economy and excluded input prices for services, imports, capital
investment, and labor. In 2015, BLS improved the construction industry input indexes by adding prices for services
inputs. At the same time, the input price index series was expanded with indexes for a small number of
manufacturing and service industries. After 2015, BLS continued its improvement efforts, and, in September 2020,
it introduced a new satellite series of input indexes. While the new satellite indexes complement official BLS
indexes already in existence, they are produced and published separately from them. The new data series
improves upon the existing input price indexes by adding prices for imported goods inputs. In addition, it

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

represents a major coverage expansion, because the official input indexes are only available for construction
industries and a limited number of manufacturing and service industries.
This article describes the methodology used to construct the new input indexes, explains their publication
structure, and provides examples of their potential uses.

Methodology
The BLS net inputs to industry price indexes presented in this article measure the average change in prices most
domestic industries pay for nearly all inputs to production, excluding capital investment and labor. As noted
previously, by tracking price changes for both domestically produced and imported inputs, these new indexes differ
from currently published BLS net inputs to industry price indexes, which account only for prices of domestically
produced inputs.
To construct an overall net inputs to industry price index, BLS calculates two separate indexes: one measuring
price change for domestically produced inputs and the other measuring price change for imported goods inputs.3
The two indexes are then aggregated into an overall input price index that measures price change for industry
inputs regardless of their country of origin. PPI commodity indexes are used to construct the domestic portion of
the overall index, and import price indexes (MPIs) are used to construct the imported portion of the index.
The first subsection below describes the use of U.S. Bureau of Economic Analysis (BEA) Input–Output (I–O) data
in establishing the set of inputs consumed by an industry. The next two subsections explain the methodologies for
calculating the price indexes for domestically produced and imported goods inputs. The final subsection presents
the methodology for combining these indexes into an overall net inputs to industry price index.

Determining inputs to an industry
To determine the set of inputs consumed by an industry, BLS relies on the BEA “Use of commodities by industries”
table (hereafter referred to as the “use table”).4 The use table provides, on an industry basis, the set and dollar
value of products consumed by each domestic industry as inputs to production. The data in the use table are
classified by I–O codes, which are very similar to NAICS codes. Importantly, the values included in the use table
represent the combined value of domestic and foreign production of the product consumed by an industry. For this
reason, the set of domestic inputs included in a given industry input index is the same as the set of imported inputs
included in that index. However, as explained below, the weights assigned to the domestically produced
commodity differ from those assigned to the imported commodity. The domestic weights reflect the relative value of
the input commodity produced in the United States, whereas the imported weights reflect the relative value of the
input commodity produced abroad.

Domestic index
The domestic portion of a net inputs to industry price index is constructed from PPI commodity indexes, which
measure price change for domestically produced goods, services, and construction products. In determining the
appropriate set of PPI commodity indexes to be included in an input price index for an industry, the PPI program
matches the industry’s use-table data to PPI codes.5 This matching results in a set of PPIs that correspond with
the products the industry consumes.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

After the set of PPIs to be included in an industry’s input price index is determined, it is necessary to construct
weights for each component PPI. These weights reflect the value of an input relative to all inputs consumed by the
industry. The gross weight for a component PPI equals the share of the total value of the commodity consumed by
the industry multiplied by the U.S. Census Bureau wherever-made value of shipments for that commodity, which
reflects the total value of the commodity’s domestic production in a given year. Assuming there are n industries
and m commodities, one can calculate the share of commodity c consumed by industry i in base period b as

where

denotes the use of commodity c by industry i in base period b; and

is the total

use of commodity c by all use-table industries in base period b.
The gross weight of commodity c in the input index for industry i at time b can then be written as

where

is the wherever-made value of shipments for commodity c in base period b.

After the gross weight of an input commodity is determined, it is converted to a net weight by removing the portion
of the commodity’s value that was produced within the industry. Net weighting removes multiple-counting bias from
the overall input price index. This bias occurs when prices from several stages of production are included in an
aggregate index.
A net weight is calculated by applying a net input ratio to the gross weight. The net input ratio is calculated by
using data from the BEA “Make of commodities by industry” table, which provides the set and dollar value of
products made by each domestic industry,6 and represents the share of the commodity produced outside the
consuming industry. The share of commodity c produced by industry i in base period b is given by

where
is the make of commodity c by industry i in base period b; and
make of commodity c by all industries in base period b.

is the total

The net input ratio of commodity c for industry i in base period b is the share of commodity c not made by industry i
and is calculated as follows:

The final net value weight for commodity c in the input index for industry i in base period b is calculated as

which can be rewritten as

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Once the products and weights for a net inputs to industry price index are determined, the index is calculated with
a modified Laspeyres formula based on standard PPI methodology.7 An approximation of the PPI aggregate price
index for month t is given by

where

is the aggregate price index in period t – 1;
is the price index for commodity c in period t – 1;

and

is the price index for commodity c in period t;
is the price index for commodity c in base period b;

is the net weight for commodity c in base period b.

Import index
The imported portion of an input price index is constructed from MPIs, which measure price change for imported
commodities and are classified by NAICS codes. As was the case with domestic inputs, determining the set of
MPIs to be included in an input price index for an industry requires converting the industry’s use-table data from I–
O codes to MPIs (based on NAICS codes). This concordance results in a set of MPIs that correspond with the
products consumed by the industry, and these MPIs are the imported components of the input price index for the
industry.
After the set of MPIs to be included in an input price index is determined, weights are constructed for the
component MPIs. The gross weight for an MPI equals the share of the total value of the commodity consumed by
the industry, as shown in equation (1), multiplied by the census import trade value of shipments for the commodity
during the base period. This calculation results in weights reflecting only the foreign-produced portion of the input
commodity’s value. The gross weight of commodity c in the input index for industry i at time t can then be written
as

where

is the value of imports for commodity c in base period b.

Unlike the domestic portion of the input index, the imported portion does not require net weighting. Because
domestic industries cannot produce imports, the share of a domestic industry’s production of the import commodity
is 0 (
gross weights.

) and the net input ratio is 1. When all net input ratios equal 1, the net weights exactly equal the

Once the products and weights for an inputs to industry price index are determined, the index is calculated with a
modified Laspeyres formula, as shown in equation (7).

Aggregating the domestic and import indexes
As noted previously, the domestic and imported input price indexes are aggregated into a total index.8 The
aggregate price index at time t is given by

U.S. BUREAU OF LABOR STATISTICS

where
t;

MONTHLY LABOR REVIEW

is the aggregate price index at time t – 1;

is the domestic price index for commodity c in period

is the domestic price index for commodity c in period t – 1;

commodity c in base period b;

is the domestic price index for

is the net weight for domestic commodity c in base period b;

foreign price index for commodity c in period t;

is the

is the foreign price index for commodity c in period t – 1;

is the foreign price index for commodity c in base period b; and
commodity c in base period b.

is the net weight for foreign

Publication structure
For each three-digit NAICS industry group, BLS publishes an aggregate input index measuring price change for
inputs (excluding capital investment and labor) consumed by the group.9 BLS also publishes separate
subaggregate indexes measuring price change for domestically produced and imported inputs consumed by the
industry group. Final breakdowns under the domestic subaggregate are published for goods, services, and
construction products. No index is produced for industry groups that do not consume a sufficient quantity of inputs
in a specific goods or service category. An example of the publication structure is presented in table 1.
Table 1. Example of a publication structure for satellite input price indexes for nonmetallic mineral product
manufacturing (NAICS 327)
Title

Code

Inputs to NAICS 327, nonmetallic mineral product manufacturing, excluding capital investment and labor
Inputs to NAICS 327, domestically produced products
Inputs to NAICS 327, domestically produced goods
Inputs to NAICS 327, domestically produced services
Inputs to NAICS 327, domestically produced maintenance and repair construction
Inputs to NAICS 327, imported goods

IN327
IN3271
IN32711
IN32712
IN32713
IN3272

Note: NAICS = North American Industry Classification System.
Source: U.S. Bureau of Labor Statistics.

The publication table includes historical index values from the first period of index calculation forward. (In most
cases, the index calculation began in December 2018.) Each month, after the release of PPI and MPI data, data
for the current period are added to the table, and data for the 4 months prior to the current period are revised.
These published data are rounded to the third decimal place.10

Data uses
This section describes four potential uses for the BLS satellite index series: industry cost analysis, price
transmission analysis, contract escalation, and deflation.

Industry cost analysis
The most straightforward use of the net inputs to industry price indexes is to measure changes in industry input
costs over time. Calculating the percent change in index levels between two periods provides a measure of the
change in an industry’s input costs, excluding those for labor and capital investment. In addition, subaggregate
indexes can be used to compare price trends for domestically produced and imported inputs. Finally, subaggregate

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

indexes within the index for domestically produced inputs can be used to compare price trends for domestically
produced goods and services inputs.
As an example, figure 1 presents input price indexes for the transportation equipment manufacturing industry
group (NAICS 336). From December 2018 through January 2020, the overall input price index rose 0.4 percent,
the index for imported inputs fell 1.1 percent, and the index for domestically produced inputs increased 1.0
percent. The inclusion of prices for imported inputs is clearly important for this industry, causing a 0.6-percentagepoint difference in the index movements over the sample period. (Without the inclusion, the index rose 1.0
percent.)

Figure 2 presents the subaggregate price index for domestically produced inputs, along with a further breakdown
for domestically produced goods and services. For transportation equipment manufacturing, the input index for
domestically produced goods fell 2.9 percent, while that for services rose 6.2 percent. Therefore, the increase in
the overall input price index can be traced to prices for domestically produced services, because prices for both
imported and domestically produced goods declined. (See figure 1.)

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

This subsection has illustrated that the new input price indexes provide a relatively detailed measure of changes in
industry costs, allowing data users to compare price trends for domestically produced and imported inputs and for
domestically produced goods and services inputs.

Price adjustments for contracting parties
BLS price index data are widely recognized as useful in price adjustment clauses, because they provide an
objective price-change measure free from possible manipulation by contracting parties. The satellite series of input
price indexes offers a new set of data that contracting parties can use in price adjustment clauses. In some cases,
contracting parties may want to make price adjustments based on either broad-level inflation or price change for a
specific product. In other instances, the parties may prefer to adjust prices on the basis of changes in overall input
costs. For these latter cases, the input price indexes may be useful by providing an objective measure of price
change for inputs purchased by specific industry groups.
There are several methods of using BLS data in price adjustment, including the “simple percentage” method, the
“adjusting a portion of the base price” method, the “index points” method, the “limits for price adjustment” method,
and the “composite indexes” method.11 An example using the composite indexes method, which is often
implemented by parties that want to adjust prices on the basis of changes in input costs, is provided below. To
implement this method, parties select a set of component price indexes that represent significant inputs and
calculate a weighted average of price change for those indexes. The weighted average is then used to adjust the
contract price. The composite indexes method often includes prices for goods inputs, services inputs, and labor.
The benefit of using a new input index is that it incorporates the near-full set of material and services inputs
consumed by an industry group. Without this index, parties would need to identify important material and services

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

inputs individually and then attempt to weight them properly, likely obtaining a less complete input measure than
the one calculated by BLS.
Table 2 presents an example of a composite price adjustment for the sale of 500 plastic containers. The
contracting parties agree to use the net inputs to industry price index for plastic and rubber products (NAICS 326),
to account for price changes in material and services inputs, and the Employment Cost Index, to account for
changes in labor input costs.12
Table 2. Example of composite price adjustment for plastic containers (NAICS 326)
Step

Inputs to NAICS 326 Employment Cost Index Composite

1: Base price (December 2018) = $1,000 per 500 plastic containers
2: 2018 index value
3: 2019 index value
4: Index relative = (2019 index)/(2018 index)
5: Assigned proportion
6: (Index relative) × (Assigned proportion)
7: Composite relative = Sum [(Index relative) x (Assigned proportion)]
8: Adjusted price = (Base price) × (Composite relative)

—
100.0
95.4
0.954
0.65
0.620
—
—

—
131.9
135.8
1.030
0.35
0.360
—
—

—
—
—
—
—
—
0.980
$980

Note: NAICS = North American Industry Classification System.
Source: Authors' calculations based on data from the U.S. Bureau of Labor Statistics.

As shown in row 1 of table 2, in the contract’s base period (December 2018), the price of 500 plastic containers is
set at $1,000. To adjust the contract price from the base period to the current period, one begins by deriving
relatives of the component indexes, dividing the 2019 index values by the 2018 index values.13 (See row 4.) Then,
the index relatives are multiplied by their respective proportions, which have been agreed to by the contracting
parties. (See row 6.) The composite relative is then calculated by summing the values from step 6. (See row 7.)
Finally, the adjusted price is calculated by multiplying the base period price by the composite relative. (See row 8.)
In this example using the composite indexes method, the base-period price of $1,000 is adjusted downward to a
new price of $980.

Price transmission analysis
The analysis of price transmission involves estimating the causal relationships between prices in a supply chain.
The new satellite net inputs to industry indexes provide data users with an opportunity to analyze price
transmission between BLS input and output price indexes for industry groups.
A rigorous price transmission analysis uses econometric time-series models to estimate the causal relationships
between prices in a supply chain.14 To be accurately estimated, these econometric models require some minimum
amount of data. In the case of the new net inputs to industry price indexes, for which most data begin in December
2018, the sample period is too short for a formal econometric analysis. For this reason, this subsection presents an
informal comparison of prices. Figure 3 displays the input and output price indexes for leather and applied product
manufacturing (NAICS 316). For comparison purposes, both indexes are rebased to equal 100 in December 2018.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

The trends in the input and output price indexes presented in figure 3 are visually similar, suggesting price
transmission between the indexes. After an initial drop in the input price index in January 2019, both the input and
output price indexes trended upward during the first quarter of 2019, declined for most of the second quarter, and
then turned up in July 2019, rising for several months. The trends diverged slightly in October 2019, as the input
price index fell while the output price index remained flat. However, for most of the final quarter of 2019, and
through January 2020, both indexes trended up. A relatively high correlation of 0.67 between the monthly percent
changes in the input and output price indexes also suggests price transmission.
Although the price transmission relationship appears to be relatively strong in leather and applied product
manufacturing, it may be weaker in other industries. This is particularly evident in cases in which wages account
for a larger share of an industry’s inputs or in which the industry is subject to frequent demand shocks. Figure 4
presents an example for a second industry, accommodation (NAICS 721), in which the relationship between input
and output prices appears to be weaker.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

While both the input and output price indexes for accommodation saw nearly identical increases over the sample
period, they did not exhibit similar turning points, and the output price index displayed much more volatility. The
output price index generally rose from December 2018 through the summer of 2019, except for a small downturn
in April, and then fell for most of the remainder of 2019 before beginning to increase again in January 2020. In
contrast, the input price index rose over almost the entire sample period, except for a small decline from
September through November 2019. The substantial difference in trends between the input and output price
indexes suggests that price transmission in accommodation is weaker than price transmission in leather and
applied product manufacturing. The correlation in monthly percent changes between the input and output price
indexes in accommodation is 0.37, which also indicates weak price transmission in that industry.
A closer examination of the output price index for accommodation over a longer period may partially explain the
relatively weak price transmission in the industry. Figure 5 presents the output price index for accommodation from
January 2012 to December 2019.

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

The output price index for accommodation appears to exhibit a pronounced seasonal pattern, with a strong peak in
the summer.15 This seasonal pattern is likely due to an increase in demand for accommodation during the summer
months. In contrast, the input index for accommodation does not appear to exhibit this type of seasonality. This
difference in seasonal patterns partially explains the relatively weak price transmission between the input and
output price indexes. Of course, the long-term trends of the indexes appear to be similar, but the available input
price data are insufficient to make this determination.
For industries that consume a substantial amount of imports as inputs, the inclusion of both domestic and imported
inputs is likely important for price transmission analysis. To illustrate this point, table 3 presents an example for
apparel manufacturing (NAICS 315), showing correlations in 1-month percent index changes (from January 2019
through April 2020) between the industry’s output price index and its input price indexes for total, domestic, and
imported inputs.
Table 3. Correlations in 1-month percent index changes between the output price index and input price
indexes for apparel manufacturing (NAICS 315)
Input price index

Relative importance

Total inputs
Domestic inputs
Imported inputs

Correlation with output index
100.0
71.1
28.9

Note: NAICS = North American Industry Classification System.
Source: Authors' calculations based on data from the U.S. Bureau of Labor Statistics

0.30
0.22
0.24

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

The correlation between the overall input price index (for both domestically produced and imported inputs) and the
output price index is approximately 0.08 points (or 36 percent) higher than the correlation between the input index
for domestically produced inputs and the output price index. Although this example is based on a short timeframe
and a limited analysis method, it illustrates that, in cases in which imports account for a relatively large portion of
industry inputs, the inclusion of imported inputs may be important for price transmission analysis.

Deflation
Deflation entails removing the effect of price changes from a revenue stream in order to separate changes in
revenue due to changes in product quantities sold from changes in revenue due to changes in prices. A revenue
stream is deflated (i.e., converted from nominal to real) with the following formula:

Applying equation (10) to a nominal revenue value converts it to a real revenue value expressed in constant
dollars from the price index’s base period. In the context of deflation, the base period is the period in which the
index equals 100.
For most revenue streams, output price indexes or consumer price indexes that correspond with the industry or
product whose revenue is being deflated are used as deflators. For some industries and products, however,
corresponding price indexes are not available. BLS may not calculate a price index for an industry or product for
two main reasons. First, the product or industry may be in scope for a BLS pricing program, but the program may
lack the resources to produce the price index. For example, PPI does not currently publish price indexes for
industries in the education sector (NAICS 611) and for the custom computer programming services industry
(NAICS 541511). Second, the product or industry may be out of scope for BLS because it has no marketed output.
For example, the temporary shelters industry (NAICS 624221), which provides short-term emergency shelter for
victims of violence and child abuse, as well as for other people in need, does not typically sell its output. Therefore,
BLS cannot calculate an output price index for temporary shelters, because there are no prices for that industry’s
output. In cases in which no output or consumer price index is available, an input index can be used as a deflator.
Table 4 presents an example of deflating U.S. Census Bureau revenue data for the social assistance industry
group (NAICS 624) with the BLS input price index for that group. BLS does not currently publish an output price
index for social assistance, because most industries within this three-digit NAICS group are out of scope for the
PPI program. In the example presented in table 4, the net inputs to industry price index is first converted from a
monthly index to a quarterly index by averaging the three monthly price indexes in each quarter. This step is
necessary because the census revenue data are only available quarterly. Next, the quarterly price index data are
rebased to equal 100 in the first quarter of 2019. This step ensures that, after the deflation is complete, the
resulting revenue will be expressed in constant first-quarter-2019 dollar values. The nominal revenue (shown in the
third data column of table 4) is then converted to real revenue with the use of equation (10).

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

Table 4. Deflation of U.S. Census Bureau data for social assistance (NAICS 624) using an input price
index, 2019
Quarter
First quarter
Second
quarter
Third quarter
Fourth
quarter

Price index Price index rebased

Nominal revenue (millions of dollars)

Real revenue (millions of dollars)

100.418

100.000

$49,431

102.145

101.720

51,385

50,516

102.423

101.997

51,194

50,192

102.346

101.920

55,954

54,900

Note: NAICS = North American Industry Classification System.
Source: Authors' calculations based on data from the U.S. Bureau of Labor Statistics and the U.S. Census Bureau.

Finally, it should be noted that the example presented in table 4 is likely oversimplified. In practice, data users
performing a deflation for social assistance may want to combine BLS input price index data with wage data,
because wages represent a substantial portion of this industry group’s inputs.16

Conclusion
In September 2020, BLS introduced a new set of satellite net inputs to industry price indexes. These
indexes measure price change for inputs (excluding capital investment and labor) consumed by most three-digit
NAICS industries and are constructed by combining PPI commodity indexes and MPIs. This article has identified a
number of potential uses for the new indexes, including industry cost analysis, contract escalation, price
transmission analysis, and deflation. BLS will be further examining this satellite data series and soliciting user
feedback on it, aiming to make it an official series in the future.
SUGGESTED CITATION

Jayson Pollock and Jonathan C. Weinhagen, "A new BLS satellite series of net inputs to industry price indexes:
methodology and uses," Monthly Labor Review, U.S. Bureau of Labor Statistics, September 2020, https://doi.org/
10.21916/mlr.2020.22.
NOTES
1 As of this article’s release, the term “satellite” is used to describe the new input price indexes because, while the indexes provide
data that expand the analytical utility of the currently published input price indexes, they are not considered an official U.S. Bureau of
Labor Statistics (BLS) output. Upon further review, BLS may release the new index series as an official output as early as 2021.
2 Input components for which BLS does not calculate price indexes are also excluded from the satellite series. Most importantly, BLS
does not calculate price indexes for approximately 28 percent of domestically produced services or any imported services. For cases
in which coverage is missing for a substantial portion of an industry’s inputs, a net input price index for that industry is not produced.
3 BLS does not publish an overall services import price index; imported services account for one-fifth of U.S. imports. Import price
index measures are available at https://www.bls.gov/mxp/.
4 The BEA use table is available at http://www.bea.gov/industry/io_annual.htm.
5 The industry input index excludes cases in which the I–O code is out of scope or not currently covered by the PPI program. BLS
also implements a cutoff rule that removes commodities accounting for less than 0.5 percent of an industry’s total inputs from the

U.S. BUREAU OF LABOR STATISTICS

MONTHLY LABOR REVIEW

industry’s input indexes. The cutoff rule substantially reduces the work required to build and maintain the net inputs to industry price
indexes, while having a negligible effect on index movements.
6 The BEA “Make of commodities by industries” table is available at http://www.bea.gov/industry/io_annual.htm.
7 For an overview of the PPI methodology, see chapter 14, “Producer prices,” Handbook of Methods (U.S. Bureau of Labor Statistics),
https://www.bls.gov/opub/hom/pdf/homch14.pdf.
8 See ibid.
9 The BLS satellite net inputs to industry price indexes for three-digit NAICS industry groups are published at https://www.bls.gov/ppi/
a-new-bls-satellite-series-inputs-to-industry-price-indexes.htm. Input indexes were not calculated if prices for a substantial portion of
an industry group’s inputs were not available.
10 Official PPIs are revised only once, 4 months after original publication. In addition, official PPI data are rounded to the first decimal
place.
11 For descriptions of these methods and for an overview of applying PPIs to price adjustment (escalation) clauses, see “Price
adjustment guide for contracting parties,” Producer Price Indexes (U.S. Bureau of Labor Statistics, 2017), https://www.bls.gov/ppi/
ppiescalation.htm.
12 The specific Employment Cost Index (ECI) used in the example is for total compensation (wages and benefits), private industry,
and goods-producing industries (database code CIU201G000000000I, https://www.bls.gov/ncs/ect/). BLS does not directly assist in
writing contracts and does not make recommendations about what data or indexes contracting parties should use.
13 Because the ECI is published on a quarterly basis, fourth-quarter values are used for the price adjustment calculation. The input
index is a monthly index, so December values are used for the calculation.
14 See, for example, Jonathan C. Weinhagen, “Price transmission: from crude petroleum to plastics products,” Monthly Labor
Review, December 2006, https://www.bls.gov/opub/mlr/2006/12/art4full.pdf.
15 Formal seasonality testing based on the U.S. Census Bureau X-12-ARIMA program indicates that this series exhibits statistically
significant seasonality.
16 The BLS ECI series is a potential source of data for measuring wages. These data can be found at https://www.bls.gov/ncs/ect/.

RELATED CONTENT

Related Articles
Measuring the substitution effect in Producer Price Index goods data: 2002–16, Monthly Labor Review, July 2020.
Comparing NAICS-based Producer Price Index industry net output data and International Price Program import data, Monthly Labor
Review, March 2018.
Price transmission within the Producer Price Index Final Demand–Intermediate Demand aggregation system, Monthly Labor Review,
August 2016.
New PPI net inputs to industry indexes, Monthly Labor Review, October 2015.

Related Subjects
Imports

Statistical methods

Producer price index

Prices

Industry studies

Full text of Monthly Labor Review : September 2020

FRASER