We test whether occupations wwhere non-routine tasks are more important are characterized by greater wage dispersion. This assumption is implied by most models of automation, yet it has not been tested earlier. By and large, the research confirms the existence of a relation, even after the inclusion of a generous set of fixed effects and occupation characteristics such as changes in relative employment and alternative measures of dispersion.
Within occupation wage dispersion and the task content of jobs
1. Within occupation wage dispersion and the task content of jobs
Within occupation wage dispersion and the task
content of jobs
Lucas van der Velde
GRAPE
Group for Research in APplied Economics
June 2017
International Economic Association
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
2. Within occupation wage dispersion and the task content of jobs
Introduction
Motivation
Theoretical concerns:
A First test on models’ assumption
(Autor et al. 2003, 2006, Acemoglu and Autor 2011, Jung and Mercenier 2014)
B Circular reasoning
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
3. Within occupation wage dispersion and the task content of jobs
Introduction
Motivation
Theoretical concerns:
A First test on models’ assumption
(Autor et al. 2003, 2006, Acemoglu and Autor 2011, Jung and Mercenier 2014)
B Circular reasoning
Policy relevance:
C Wage dispersion within occupation is greater than between
D Within occupation wage dispersion can increase in future
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
4. Within occupation wage dispersion and the task content of jobs
Introduction
Assumption of task content models
Productivity in different tasks and individual ability
(Jung and Mercenier 2014, Cortes 2016):
0 <
∂ lnφm(z)
∂z
<
∂ lnφr (z)
∂z
<
∂ lnφnr (z)
∂z
,
where
→ z indicates worker ability
→ φi (z) worker’s productivity in task i
Variations over this subject offered in Autor et al. (2003, 2006), Beaudry et al. (2016)
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
5. Within occupation wage dispersion and the task content of jobs
Introduction
Importance of within occupation wage dispersion
Figure: Sources of wage inequality in selected European countries
0 .1 .2 .3
Value of the index
BGR
LVA
GBR
LTU
EST
POL
CYP
FRA
SVK
CZE
LUX
3 digit ISCO codes
0 .1 .2 .3
Value of the index
ROM
PRT
DEU
ESP
ITA
HUN
NLD
GRC
BEL
FIN
NOR
SWE
2 digit ISCO codes
Within occupations Between occupations
Source: Structure of Earning Survey of the EU, 2010.
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
6. Within occupation wage dispersion and the task content of jobs
Introduction
Hypothesis
Hypothesis
Within occupation wage dispersion is positively correlated with the
non-routine task content of the occupation, ceteris paribus
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
7. Within occupation wage dispersion and the task content of jobs
Data and method
Data: O*NET
Task content of occupations
Occupational Network (O*NET) data from 2010
Data collected from US workers, available at occupation level
Information on importance and frequency of tasks
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
8. Within occupation wage dispersion and the task content of jobs
Data and method
Data: O*NET
Task content of occupations
Occupational Network (O*NET) data from 2010
Data collected from US workers, available at occupation level
Information on importance and frequency of tasks
Five tasks (as in e.g. Autor et al. 2003, Acemoglu and Autor 2011)
Routine: cognitive and manual
Non-routine: cognitive, interpersonal and manual
Routine task intensity index RTI = routine − non-routine
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
9. Within occupation wage dispersion and the task content of jobs
Data and method
Data: O*NET
Task content of occupations
Occupational Network (O*NET) data from 2010
Data collected from US workers, available at occupation level
Information on importance and frequency of tasks
Five tasks (as in e.g. Autor et al. 2003, Acemoglu and Autor 2011)
Routine: cognitive and manual
Non-routine: cognitive, interpersonal and manual
Routine task intensity index RTI = routine − non-routine
Applied to EU countries before
(e.g. Goos and Manning 2007, Goos et al. 2014, Lewandowski et al. 2016)
Independent measure for each country.
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
10. Within occupation wage dispersion and the task content of jobs
Data and method
Data: O*NET
Task content of occupations
Occupational Network (O*NET) data from 2010
Data collected from US workers, available at occupation level
Information on importance and frequency of tasks
Five tasks (as in e.g. Autor et al. 2003, Acemoglu and Autor 2011)
Routine: cognitive and manual
Non-routine: cognitive, interpersonal and manual
Routine task intensity index RTI = routine − non-routine
Applied to EU countries before
(e.g. Goos and Manning 2007, Goos et al. 2014, Lewandowski et al. 2016)
Independent measure for each country.
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
11. Within occupation wage dispersion and the task content of jobs
Data and method
Data: EU-SES
Structure of Earnings Survey
⇒ Employee-employer matched data: years 2002, 2006 and 2010
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
12. Within occupation wage dispersion and the task content of jobs
Data and method
Data: EU-SES
Structure of Earnings Survey
⇒ Employee-employer matched data: years 2002, 2006 and 2010
⇒ Why...?
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
13. Within occupation wage dispersion and the task content of jobs
Data and method
Data: EU-SES
Structure of Earnings Survey
⇒ Employee-employer matched data: years 2002, 2006 and 2010
⇒ Why...?
1. Large sample sizes
2. Reliable information on hourly wages and occupations.
3. Comparable information for EU countries.
Descriptive statistics
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
14. Within occupation wage dispersion and the task content of jobs
Data and method
Empirical approach
Specification
wage dispersionj,t = α0 + βRTIj,t + γD +
where
Wage dispersion in occupation j in t→ e.g. 90th to 10th percentiles
β is the coefficient of interest → H1: β is negative and significant
D represents a set of controls
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
15. Within occupation wage dispersion and the task content of jobs
Results
Wage dispersion and RTI
wage dispersionj,t = α0 + βRTIj,t + γD +
Measure of dispersion 90/10 90/50 50/10
Notes: Robust standard errors used in all regressions.
*,**,*** indicate significance at the 10%, 5% and 1%
level. All regressions include country and year fixed ef-
fects.
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
16. Within occupation wage dispersion and the task content of jobs
Results
Wage dispersion and RTI
wage dispersionj,t = α0 + βRTIj,t + γD +
Measure of dispersion 90/10 90/50 50/10
Unconditional
Wages
Notes: Robust standard errors used in all regressions.
*,**,*** indicate significance at the 10%, 5% and 1%
level. All regressions include country and year fixed ef-
fects.
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
17. Within occupation wage dispersion and the task content of jobs
Results
Wage dispersion and RTI
wage dispersionj,t = α0 + βRTIj,t + γD +
Measure of dispersion 90/10 90/50 50/10
Unconditional
Wages
Conditional
Wages
Notes: Robust standard errors used in all regressions.
*,**,*** indicate significance at the 10%, 5% and 1%
level. All regressions include country and year fixed ef-
fects.
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
18. Within occupation wage dispersion and the task content of jobs
Results
Wage dispersion and RTI
wage dispersionj,t = α0 + βRTIj,t + γD +
Measure of dispersion 90/10 90/50 50/10
Unconditional β -0.09***
Wages R2
0.48
Conditional β -0.07***
Wages R2
0.54
Notes: Robust standard errors used in all regressions.
*,**,*** indicate significance at the 10%, 5% and 1%
level. All regressions include country and year fixed ef-
fects.
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
19. Within occupation wage dispersion and the task content of jobs
Results
Wage dispersion and RTI
wage dispersionj,t = α0 + βRTIj,t + γD +
Measure of dispersion 90/10 90/50 50/10
Unconditional β -0.09*** -0.04*** -0.04***
Wages R2
0.48 0.42 0.39
Conditional β -0.07*** -0.04*** -0.03**
Wages R2
0.54 0.48 0.47
Notes: Robust standard errors used in all regressions.
*,**,*** indicate significance at the 10%, 5% and 1%
level. All regressions include country and year fixed ef-
fects.
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
20. Within occupation wage dispersion and the task content of jobs
Results
Which tasks are driving the results?
wage dispersionj,t = α0 +
5
i=1
βi Ti,j,t + γD +
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
21. Within occupation wage dispersion and the task content of jobs
Results
Which tasks are driving the results?
wage dispersionj,t = α0 +
5
i=1
βi Ti,j,t + γD +
90/10 90/50 50/10
Uncond. Cond. Uncond. Cond. Uncond. Cond.
NR Cog. 0 0.07*** -0.01 0.04*** 0.01 0.04***
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
22. Within occupation wage dispersion and the task content of jobs
Results
Which tasks are driving the results?
wage dispersionj,t = α0 +
5
i=1
βi Ti,j,t + γD +
90/10 90/50 50/10
Uncond. Cond. Uncond. Cond. Uncond. Cond.
NR Cog. 0 0.07*** -0.01 0.04*** 0.01 0.04***
NR. Per. 0.07*** 0.04*** 0.04*** 0.02*** 0.03*** 0.02***
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
23. Within occupation wage dispersion and the task content of jobs
Results
Which tasks are driving the results?
wage dispersionj,t = α0 +
5
i=1
βi Ti,j,t + γD +
90/10 90/50 50/10
Uncond. Cond. Uncond. Cond. Uncond. Cond.
NR Cog. 0 0.07*** -0.01 0.04*** 0.01 0.04***
NR. Per. 0.07*** 0.04*** 0.04*** 0.02*** 0.03*** 0.02***
NR. Man. -0.02 -0.01 -0.01 -0.01 -0.01 0.00
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
24. Within occupation wage dispersion and the task content of jobs
Results
Which tasks are driving the results?
wage dispersionj,t = α0 +
5
i=1
βi Ti,j,t + γD +
90/10 90/50 50/10
Uncond. Cond. Uncond. Cond. Uncond. Cond.
NR Cog. 0 0.07*** -0.01 0.04*** 0.01 0.04***
NR. Per. 0.07*** 0.04*** 0.04*** 0.02*** 0.03*** 0.02***
NR. Man. -0.02 -0.01 -0.01 -0.01 -0.01 0.00
R. Cog. 0.05*** 0.01** 0.01*** 0 0.03*** 0.01***
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
25. Within occupation wage dispersion and the task content of jobs
Results
Which tasks are driving the results?
wage dispersionj,t = α0 +
5
i=1
βi Ti,j,t + γD +
90/10 90/50 50/10
Uncond. Cond. Uncond. Cond. Uncond. Cond.
NR Cog. 0 0.07*** -0.01 0.04*** 0.01 0.04***
NR. Per. 0.07*** 0.04*** 0.04*** 0.02*** 0.03*** 0.02***
NR. Man. -0.02 -0.01 -0.01 -0.01 -0.01 0.00
R. Cog. 0.05*** 0.01** 0.01*** 0 0.03*** 0.01***
R. Man. -0.01 0 -0.01 0.01 0 -0.01
N 1862 1862 1862 1862 1862 1862
R2 0.65 0.59 0.55 0.51 0.55 0.52
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
26. Within occupation wage dispersion and the task content of jobs
Results
Alternative explanations I
Reallocation induced wage dispersion
Autor et al. (2003), Acemoglu and Autor (2011), Jung and Mercenier (2014)
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
27. Within occupation wage dispersion and the task content of jobs
Results
Alternative explanations I
Reallocation induced wage dispersion
Autor et al. (2003), Acemoglu and Autor (2011), Jung and Mercenier (2014)
wage dispersionj,t = α0 + β1RTIj,t + β2Occ changet,t−5 + γD +
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
28. Within occupation wage dispersion and the task content of jobs
Results
Alternative explanations I
Reallocation induced wage dispersion
Autor et al. (2003), Acemoglu and Autor (2011), Jung and Mercenier (2014)
wage dispersionj,t = α0 + β1RTIj,t + β2Occ changet,t−5 + γD +
Unconditional
Conditional
−.03 −.02 −.01 0 .01 .02
90/10 90/50 50/10
RTI
Unconditional
Conditional
−1 0 1 2 3
90/10 90/50 50/10
Emp. Change
Notes: Coefficients and 95% CI graphed. Regressions include year and country fixed effects
and robust standard errors.
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
29. Within occupation wage dispersion and the task content of jobs
Results
Alternative explanations II
Winner-takes-all markets and RTI
Brynjolfsson and McAfee (2016)
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
30. Within occupation wage dispersion and the task content of jobs
Results
Alternative explanations II
Winner-takes-all markets and RTI
Brynjolfsson and McAfee (2016)
wage dispersionj,t = α0 + β1RTIj,t + β2S1 S10 + γD +
Unconditional wages
Conditional wages
−.1 −.08 −.06 −.04 −.02
90/10 90/50 50/10
RTI
Unconditional
Conditional
.005 .01 .015 .02 .025
90/10 90/50 50/10
Share ratio
Notes: Coefficients and 95% CI graphed. Regressions include year and country fixed effects
and robust standard errors.
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
31. Within occupation wage dispersion and the task content of jobs
Results
Conclusions
1. New test of RBTC hypothesis predictions
Within occupation wage dispersion is greater in occupations with
more non-routine content
2. Confirmed and robust
Alternative channels do not affect conclusions
Other robustness checks: nonlinearities in RTI, more disaggregated
occupations, alternative definitions of occupational change, task
content over regions → results are consistent
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
32. Within occupation wage dispersion and the task content of jobs
Results
Conclusions
1. New test of RBTC hypothesis predictions
Within occupation wage dispersion is greater in occupations with
more non-routine content
2. Confirmed and robust
Alternative channels do not affect conclusions
Other robustness checks: nonlinearities in RTI, more disaggregated
occupations, alternative definitions of occupational change, task
content over regions → results are consistent
Implications for theory
→ Data confirm characterization of routine/ non-routine tasks
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
33. Within occupation wage dispersion and the task content of jobs
Results
Thank you for your attention
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
34. Within occupation wage dispersion and the task content of jobs
Results
References I
Acemoglu, D. and Autor, D.: 2011, Skills, tasks and technologies: Implications for
employment and earnings, Handbook of Labor Economics 4, 1043–1171.
Autor, D., Katz, L. F. and Kearney, M. S.: 2006, The polarization of the US labor
market, American Economic Review 96(2), 189–194.
Autor, D., Levy, F. and Murnane, R. J.: 2003, The skill content of recent
technological change: An empirical exploration, Quarterly Journal of Economics
118(4), 1279–1333.
Beaudry, P., Green, D. A. and Sand, B. M.: 2016, The great reversal in the demand
for skill and cognitive tasks, Journal of Labor Economics 34(S1), S199–S247.
Cortes, G. M.: 2016, Where have the middle-wage workers gone? A study of
polarization using panel data, Journal of Labor Economics 34(1), 63–105.
Goos, M. and Manning, A.: 2007, Lousy and lovely jobs: The rising polarization of
work in Britain, Review of Economics and Statistics 89(1), 118–133.
Goos, M., Manning, A. and Salomons, A.: 2014, Explaining job polarization:
Routine-biased technological change and offshoring, American Economic Review
104(8), 2509–2526.
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
35. Within occupation wage dispersion and the task content of jobs
Results
References II
Jung, J. and Mercenier, J.: 2014, Routinization-biased technical change and
globalization: Understanding labor market polarization, Economic Inquiry
52(4), 1446–1465.
Lewandowski, P., Hardy, W. and Keister, R.: 2016, Technology or upskilling? Trends
in the task tomposition of jobs in Central and Eastern Europe, IBS Working Papers
(01/2016).
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
36. Within occupation wage dispersion and the task content of jobs
Appendix
Descriptive statistics: EU-SES
Sample description
Year 2002 2006 2010
Avg. Nocc 11 747 (18 421) 18 876 (51 498) 12 491 (23 744)
Median Nocc 5 120 6 627 4 926
- Wage inequality
p90/p10 3.73 (1.13) 3.72 (1.14) 3.61 (0.92)
p90/p50 2.10 (0.41) 2.07 (0.35) 2.06 (0.33)
p50/p10 1.75 (0.3) 1.77 (0.3) 1.73 (0.26)
Notes: Standard deviations presented in parentheses. Sample includes all EU members except
Cyprus and Luxembourg.
Back
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
37. Within occupation wage dispersion and the task content of jobs
Appendix
Regional effects I
Figure: Within occupation wage inequality and task content in regions
0
.05
.1
.15
.2
Withinocc.wageinequality
−.5 0 .5 1
Non routine cognitive content
0
.05
.1
.15
.2
Withinocc.wageinequality
−.2 0 .2 .4 .6 .8
Routine cognitive content
0
.05
.1
.15
.2
Withinocc.wageinequality
−.5 0 .5 1
Routine manual content
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
38. Within occupation wage dispersion and the task content of jobs
Appendix
Regional effects II
Y Within occupation wage inequality
X NR-Cog R-Cog R-Manual
β 0.03*** 0.11*** 0.02 -0.08** -0.03*** -0.07***
SE (0.01) (0.03) (0.01) (0.05) (0.01) (0.03)
N 202 202 202 202 202 202
R2 0.02 0.65 0.01 0.61 0.02 0.65
F.E. Yes Yes Yes
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs
39. Within occupation wage dispersion and the task content of jobs
Appendix
Regional effects II
Y Within occupation wage inequality
X NR-Cog R-Cog R-Manual
β 0.03*** 0.11*** 0.02 -0.08** -0.03*** -0.07***
SE (0.01) (0.03) (0.01) (0.05) (0.01) (0.03)
N 202 202 202 202 202 202
R2 0.02 0.65 0.01 0.61 0.02 0.65
F.E. Yes Yes Yes
Y Within occupation wage inequality/ total inequality
β -0.00 0.10* 0.04 -0.05 0.04** -0.04
SE (0.03) (0.06) (0.03) (0.11) (0.02) (0.05)
N 202 202 202 202 202 202
R2 0.00 0.74 0.01 0.73 0.01 0.73
F.E. Yes Yes Yes
Notes: F.E. indicate regional and year fixed effects. Dependent variable is the average value
of task i in region j in period t. Robust standard errors in parentheses. *,**,*** indicate
significance at the 10%, 5% and 1% level respectively.
van der Velde GRAPE Group for Research in APplied Economics
Within occupation wage dispersion and the task content of jobs