Presentation: U.S. Wage Puzzles: False Trails, Clues and ...2018/01/04 · 1st (Lowest) 2nd 3rd 4th 5th (Highest) 1998:Q1– 2001:Q1 2015:Q2– 2018:Q2 Real Average Hourly Earnings

U.S. Wage Puzzles: False Trails, Clues and Answers

Jason FurmanHarvard Kennedy School

PIIE

Peterson Institute for International Economics | 1750 Massachusetts Ave., NW | Washington, DC 20036

Peterson Institute for International EconomicsWashington, DCOctober 4, 2018

Many measures of wage growth, but a (generally) consistent trend

Other wage measures: Atlanta Fed wage growth tracker, ECI: wages and salaries of civilian workers, ECEC: wages and salaries of civilian workers, ECI: wages and salaries of private industry workers, ECEC: wages and salaries of private industry workers, median usual weekly earnings, average hourly earnings: total private, average hourly earnings: production and nonsupervisory workers, average weekly earnings: total private, average weekly earnings: production and nonsupervisory workers.Source: Bureau of Labor Statistics; Current Population Survey, Merged Outgoing Rotation Groups; IPUMS CPS; Goldman Sachs; Federal Reserve Bank of Atlanta; Haver Analytics; author’s calculations.

Many different measures of wage growth show varying numbers due to different concepts or statistical noise.

My focus will mostly be on average hourly earnings for workers age 25-54.-1

0

1

2

3

4

5

6

7

1990 1995 2000 2005 2010 2015

Average Hourly Earnings: age25-54Goldman Sachs Wage Tracker

Nominal Wage GrowthPercent Change, Year Ago

-1

0

1

2

3

4

5

6

7

1990 1995 2000 2005 2010 2015



WAGE PUZZLE #1Why is wage growth slower today than in the

late 1990s?

Annual growth in average hourly earnings for age 25-54:

1998:Q1-2001:Q1: 4.8%

2015:Q2-2018:Q2: 3.6%

Nominal wage growth is 1.2 percentage point slower now than in the late 1990s.


Hypothesis 1: InflationA temporary increase in inflation due to oil

prices explains much of this year’s slowdown

Note: Shaded area represents real wage growth. Red shading indicates that real wage growth was negative. Source: Bureau of Labor Statistics; Current Population Survey, Merged Outgoing Rotation Groups; IPUMS CPS; Bureau of Economic Analysis; Haver Analytics; author’s calculations.

Recent nominal wage growth has largely been undone by rising inflation, but that is likely to be a temporary phenomenon assuming oil prices stabilize and/or fall.

Nominal Wage

Growth

PCE Price Inflation

-3

-2

-1

0

1

2

3

4

5

6

7

1990 1995 2000 2005 2010 2015

Average Hourly Earnings Growth, Prime-age (age 25-54) Wage and Salary Workers

Four-Quarter Percent Change

Hypothesis 1: InflationLower trend inflation explains part of the

slower growth of nominal wages

Note: Nominal wages are deflated by PCE price inflation. Source: Bureau of Labor Statistics; Current Population Survey, Merged Outgoing Rotation Groups; IPUMS CPS; Bureau of Economic Analysis; Haver Analytics; author’s calculations.

Instead of a 1.2 p.p.slowdown in nominal wages relative to the late 1990s, we only need to explain a 0.8 p.p. slowdown in real wages.

2018:Q21998:Q1–2001:Q1

2.8

2015:Q2–2018:Q2

2.0

-2

-1

0

1

2

3

4

1992 1997 2002 2007 2012 2017

Real Average Hourly Earnings Growth, Prime-age (age 25-54) Wage and Salary Workers

Percent Change, Trailing Three-Year Average (Annual Rate)

Hypothesis 2: InequalityDoes not explain the slower wage growth than

in the late 1990s

Note: Top-coded earnings are adjusted following Lemieux (2006). Excludes observations with hourly earnings below $0.50 or above $100 in 1989 dollars as deflated by the CPI-U-RS. Nominal wages are deflated by PCE price inflation. Source: Bureau of Labor Statistics; Haver Analytics; Current Population Survey, Merged Outgoing Rotation Groups; IPUMS CPS; Bureau of Economic Analysis; author's calculations.

0

1

2

3

4

1st (Lowest) 2nd 3rd 4th 5th (Highest)

1998:Q1–2001:Q1

Real Average Hourly Earnings Growth by Quintile, Prime-age (age 25-54) Wage and Salary Workers

Percent Change, Annual Rate

Quintile


in the late 1990s


0

1

2

3

4


1998:Q1–2001:Q1

2015:Q2–2018:Q2



Quintile


in the late 1990s


In the latest period wage growth at the bottom is higher than in the late 1990s and is higher than it has been at the top recently.

0

1

2

3

4


1998:Q1–2001:Q1

2015:Q2–2018:Q2



Quintile

Hypothesis 2: InequalityInequality was rising more rapidly in the 1980s

and 1990s than it has recently

Note: CEO compensation based on options realized.Source: Congressional Budget Office; Mishel and Schieder (2018); author’s calculations.

2014

0

3

6

9

12

15

18

1979 1989 1999 2009

Top 1% Share of After-tax Income, Non-elderly-headed Households

Percent

0

50

100

150

200

250

300

350

400

1980 1990 2000 2010

CEO-to-Worker Compensation RatioRatio

2017

Hypothesis 3: SlackMany measures of slack are lower today than in the late 1990s—except employment rates

Note: U5.5 is the number of unemployed, plus the number marginally attached to the labor force, plus 0.5 times the number working part-time for economic reasons as a share of the sum of the labor force and the number marginally attached to the labor force. Source: Bureau of Labor Statistics; Haver Analytics; author’s calculations.

Labor markets are tighter now according to some measures, but not others.

2001:Q1 2018:Q2 ChangeShort-term Unemployment Rate 3.7 3.1 -0.7

Unemployment Rate 4.2 3.9 -0.3

U5.5 6.2 6.2 0.1

Prime-age (25-54) Nonemployment Rate 18.7 20.8 2.1

Measures of Labor Marked Slack

Hypothesis 3: SlackEmployment rates have major trends unrelated

to the business cycle

Source: Bureau of Labor Statistics; Haver Analytics; author’s calculations.

2018:Q2

75

80

85

90

95

100

1948 1958 1968 1978 1988 1998 2008 2018

Prime-age (age 25-54) Male Employment-Population Ratio

Percent

1948-2007 Trend

30

35

40

45

50

55

60

65

70

75

80

1948 1958 1968 1978 1988 1998 2008 2018

Prime-age (age 25-54) Female Employment-Population Ratio

Percent

2018:Q2

1948-2000 Trend

2000-2007Trend

2018:Q2

75

80

85

90

95

100

1948 1958 1968 1978 1988 1998 2008 2018

Prime-age (age 25-54) Male Employment-Population Ratio

Percent

1948-2007 Trend

Hypothesis 3: SlackEmployment rates have major trends unrelated

to the business cycle


Did the economy really have 2 p.p.

more of slack in 2000 than in 1979?

30

35

40

45

50

55

60

65

70

75

80

1948 1958 1968 1978 1988 1998 2008 2018

Prime-age (age 25-54) Female Employment-Population Ratio

Percent

2018:Q2

1948-2000 Trend

2000-2007Trend

Hypothesis 4: Productivity growthProductivity slowdown more than explains the

difference


With lower productivity growth, we would expect wage growth to be lower—by as much as 1.8 percentage points.

1996:Q1–2001:Q1

2.8

2013:Q2–2018:Q2

1.0

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1990 1995 2000 2005 2010 2015

Productivity Growth, Nonfarm Business SectorPercent Change, Trailing Five-Year Average (Annual Rate)

Hypothesis 4: Productivity growthMacro and micro evidence both find

productivity growth is tied to wage growth

Source: Economic Policy Institute; Bureau of Economic Analysis; Bureau of Labor Statistics; Haver Analytics; author’s calculations.

Net Productivity (PCE Price

Index Deflator)

Average Compensation

(PCE Price Index Deflator)

2016

-20

0

20

40

60

80

100

1973 1978 1983 1988 1993 1998 2003 2008 2013

Productivity Growth vs. Average Compensation Growth

Percent Change Since 1973

Net Productivity (Net GDP

Price Index Deflator)

2016

Median Compensation

(CPI-U-RS Deflator)

-20

0

20

40

60

80

100

1973 1978 1983 1988 1993 1998 2003 2008 2013

Productivity Growth vs. Median Compensation Growth

Percent Change Since 1973

Mean compensation very closely tied to mean productivity growth if you use the same deflators/concepts for both. (Note—median still disconnected from mean, a distinct issue that has not been the important factor in the recent data.)

0

1

2

3

4

5

6

2015 2016 2017 2018



WAGE PUZZLE #2Why hasn’t wage growth picked up more in the

last few years?

Measures of nominal wage growth are only slightly faster over the previous year than they were three years ago


Hypothesis 1: InflationCompounds the puzzle as real wage

acceleration < nominal wage acceleration

Source: Bureau of Labor Statistics; Current Population Survey, Merged Outgoing Rotation Groups; IPUMS CPS; Bureau of Economic Analysis; Federal Reserve Bank of Cleveland; University of Michigan; Federal Reserve Bank of Philadelphia; Haver Analytics; author’s calculations.

Inflation has increased in the last few years, compounding the puzzle—real wage growth has has been effectively unchanged.0.0

0.5

1.0

1.5

2.0

2.5

AverageHourly

Earnings,Prime-age

(25-54)

PCEInflation

Core PCEInflation

TrimmedMean CPI

Univ. ofMichiganExpectedInflation,

Next Year

SPFExpected

CPIInflation,

Next Year

Change in Measures of Inflation: 2015:Q2 to 2018:Q2Percentage Points

Hypothesis 2: SlackCompounds the puzzle because labor markets have tightened, especially broader measures

Note: U5.5 is the number of unemployed, plus the number marginally attached to the labor force, plus 0.5 times the number working part-time for economic reasons as a share of the sum of the labor force and the number marginally attached to the labor force. Source: Bureau of Labor Statistics; Haver Analytics; author’s calculations.

Labor markets are much tighter now than three years ago, so looking at slack also deepens the puzzle.

Broader measures of slack have tightened even more than narrower measures have.

2015:Q2 2018:Q2 ChangeShort-term Unemployment Rate 3.9 3.1 -0.8

Unemployment Rate 5.4 3.9 -1.5

U5.5 8.6 6.2 -2.4

Prime-age (25-54) Nonemployment Rate 22.7 20.8 -2.0

Measures of Labor Marked Slack

Hypothesis 2: SlackRemember what matters is “fuller

employment” not “full employment”

Hypothesis 2: SlackRemember what matters is “fuller

employment” not “full employment”

Hypothesis 3: ProductivitySlowdown only understood and priced in with

a lag could help explain Puzzle #2

Source: Federal Reserve Bank of Philadelphia; Haver Analytics.

Falling expectations for long-term growth indicate that expectations for trend productivity growth have been revised down over time.

2018

1.5

2.0

2.5

3.0

3.5

2000 2005 2010 2015 2020

SPF: Projected Real GDP Growth Rate, Next 10 YearsPercent Change, Annual Rate

Takeaways from the wage puzzles

Explanations Wage Puzzle #1:Slower than Late 1990s

Wage Puzzle #2:Has Not Picked Up in Last 3 Years

Inflation Partly explains: Trend inflation lower so nominal wage growth lower.

Deepens the puzzle: Trend inflation has picked up.

Inequality Deepens the puzzle: Inequality was increasing much more in the late 1990s and is even decreasing in some dimensions now.

Does not explain: No evidence inequality/institutions has changed in the last few years.

Productivity More than explains: Much slower productivity growth leads to much slower wage growth.

Possible explanation: Downward revisions to productivity expectations could cause downward pressure on wage growth.

Slack Does not explain: Slack similar/tighter than late 1990s (except EPOP).

Deepens the puzzle: Slack much lower than 3 years ago, especially for broader measures.

U.S. Wage Puzzles: False Trails, Clues and Answers

Jason FurmanHarvard Kennedy School

PIIE

Peterson Institute for International Economics | 1750 Massachusetts Ave., NW | Washington, DC 20036

Peterson Institute for International EconomicsWashington, DCOctober 4, 2018

Documents

Presentation: U.S. Wage Puzzles: False Trails, Clues and ...2018/01/04 · 1st (Lowest) 2nd 3rd 4th 5th (Highest) 1998:Q1– 2001:Q1 2015:Q2– 2018:Q2 Real Average Hourly Earnings