building a productive workforce: the role of structured ...€¦ · introductiondata and summary...
TRANSCRIPT
Building a productive workforce:the role of structured management practices
Chris CornwellUniversity of Georgia
Ian SchmutteUniversity of Georgia
Daniela ScurMIT Sloan
ESCoE, NIESRApril 2019
Introduction Data and summary stats Results Conclusion
Introduction
Big picture question: how do firms build productive workforces?
I Optimize selection (sorting) of employees.
I Set optimal incentives for hired employees.
I Minimize costly turnover.
We know that:
I Managers have discretion, and thus information/agency problem arise.Hoffman et al 2018, Oyer and Schaefer 2010
I Firms with more structured practices are more productive.Bloom et al 2013, 2018, Gosnell et al 2016, Bruhn et al 2018
I Team production and substitution is important for productivity.Jager 2016, Hensvik and Rosenqvist 2019
I Assortative matching matters Abowd et al 1999, Card et al 2013, Sorkin 2017
... but we don’t have much direct evidence from inside the black box.Bandiera et al 2015, Bender et al 2016
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 2 / 24
Introduction Data and summary stats Results Conclusion
Introduction
Big picture question: how do firms build productive workforces?
I Optimize selection (sorting) of employees.
I Set optimal incentives for hired employees.
I Minimize costly turnover.
We know that:
I Managers have discretion, and thus information/agency problem arise.Hoffman et al 2018, Oyer and Schaefer 2010
I Firms with more structured practices are more productive.Bloom et al 2013, 2018, Gosnell et al 2016, Bruhn et al 2018
I Team production and substitution is important for productivity.Jager 2016, Hensvik and Rosenqvist 2019
I Assortative matching matters Abowd et al 1999, Card et al 2013, Sorkin 2017
... but we don’t have much direct evidence from inside the black box.Bandiera et al 2015, Bender et al 2016
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 2 / 24
Introduction Data and summary stats Results Conclusion
Introduction
Big picture question: how do firms build productive workforces?
I Optimize selection (sorting) of employees.
I Set optimal incentives for hired employees.
I Minimize costly turnover.
We know that:
I Managers have discretion, and thus information/agency problem arise.Hoffman et al 2018, Oyer and Schaefer 2010
I Firms with more structured practices are more productive.Bloom et al 2013, 2018, Gosnell et al 2016, Bruhn et al 2018
I Team production and substitution is important for productivity.Jager 2016, Hensvik and Rosenqvist 2019
I Assortative matching matters Abowd et al 1999, Card et al 2013, Sorkin 2017
... but we don’t have much direct evidence from inside the black box.Bandiera et al 2015, Bender et al 2016
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 2 / 24
Introduction Data and summary stats Results Conclusion
Introduction
Big picture question: how do firms build productive workforces?
I Optimize selection (sorting) of employees.
I Set optimal incentives for hired employees.
I Minimize costly turnover.
We know that:
I Managers have discretion, and thus information/agency problem arise.Hoffman et al 2018, Oyer and Schaefer 2010
I Firms with more structured practices are more productive.Bloom et al 2013, 2018, Gosnell et al 2016, Bruhn et al 2018
I Team production and substitution is important for productivity.Jager 2016, Hensvik and Rosenqvist 2019
I Assortative matching matters Abowd et al 1999, Card et al 2013, Sorkin 2017
... but we don’t have much direct evidence from inside the black box.Bandiera et al 2015, Bender et al 2016
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 2 / 24
Introduction Data and summary stats Results Conclusion
Introduction
Big picture question: how do firms build productive workforces?
I Optimize selection (sorting) of employees.
I Set optimal incentives for hired employees.
I Minimize costly turnover.
We know that:
I Managers have discretion, and thus information/agency problem arise.Hoffman et al 2018, Oyer and Schaefer 2010
I Firms with more structured practices are more productive.Bloom et al 2013, 2018, Gosnell et al 2016, Bruhn et al 2018
I Team production and substitution is important for productivity.Jager 2016, Hensvik and Rosenqvist 2019
I Assortative matching matters Abowd et al 1999, Card et al 2013, Sorkin 2017
... but we don’t have much direct evidence from inside the black box.Bandiera et al 2015, Bender et al 2016
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 2 / 24
Introduction Data and summary stats Results Conclusion
This paper: management and building a workforce
Main research question: What is the role of structured management practices inbuilding a productive workforce?
What we use: New matched data from Brazil
I Linked EE data: RAIS 2003-2013
I Firm management practices: WMS 2008, 2013
I Firm productivity: Pesquisa Industrial Anual (PIA) 2003-2013
What we do:
1 Identify managers and production workers from occupation codes.
2 Estimate (and rank) AKM person effects for managers and prod. workers.
3 Document relationship between management structure, worker selection,and productivity
4 Describe movements of workers in/out of structured management firms
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 3 / 24
Introduction Data and summary stats Results Conclusion
This paper: management and building a workforce
Main research question: What is the role of structured management practices inbuilding a productive workforce?
What we use: New matched data from Brazil
I Linked EE data: RAIS 2003-2013
I Firm management practices: WMS 2008, 2013
I Firm productivity: Pesquisa Industrial Anual (PIA) 2003-2013
What we do:
1 Identify managers and production workers from occupation codes.
2 Estimate (and rank) AKM person effects for managers and prod. workers.
3 Document relationship between management structure, worker selection,and productivity
4 Describe movements of workers in/out of structured management firms
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 3 / 24
Introduction Data and summary stats Results Conclusion
This paper: management and building a workforce
Main research question: What is the role of structured management practices inbuilding a productive workforce?
What we use: New matched data from Brazil
I Linked EE data: RAIS 2003-2013
I Firm management practices: WMS 2008, 2013
I Firm productivity: Pesquisa Industrial Anual (PIA) 2003-2013
What we do:
1 Identify managers and production workers from occupation codes.
2 Estimate (and rank) AKM person effects for managers and prod. workers.
3 Document relationship between management structure, worker selection,and productivity
4 Describe movements of workers in/out of structured management firms
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 3 / 24
Introduction Data and summary stats Results Conclusion
Overview of findings
1 Worker and manager quality are positively correlated with TFP.I Replicated result from Bender et al 2018.
What is the role of personnel management in building the workforce?
More structured management is associated with:
2 Positive recruitment: hire a larger share of their new recruits from the topof the distribution of worker quality.
3 Worker matching and separation: fire less often and more selectively.
4 Retention practices: consistently retain a high share of the best workers.
5 Different practices are linked to selectivity across levels of the hierarchy.
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 4 / 24
Introduction Data and summary stats Results Conclusion
1 Introduction
2 Data and summary statsWorkerManagementProductivity
3 Results
4 Conclusion
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 5 / 24
Introduction Data and summary stats Results Conclusion
Worker data: Brazilian employment records (RAIS)
I Collected from employers to administer social security payments.
I Formal-sector jobs (∼50 million per year).
I Includes information on:I worker characteristics: education, experience, race, sexI job characteristics: occupation, wage, hours, tenure,I job spell information: reason for separation, date of hireI employer characteristics: industry, legal structure, size, location
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 6 / 24
Introduction Data and summary stats Results Conclusion
Summary statistics: workers
Mean Median Min Max SD N
Worker characteristicsAge of workers (RAIS) 33.05 32.7 21.0 53.0 (3.75) 961Employee tenure, weeks (RAIS) 43.98 39.8 2.9 213.7 (22.12) 961Hourly wage, BRL Reais (RAIS) 11.24 8.3 2.5 159.7 (10.61) 961Monthly earnings, BRL Reais (RAIS) 2079.36 1530.3 463.4 30120.6 (1931.22) 961
Share of female workers, total (WMS) 0.30 0.3 0.0 1.0 (0.24) 480Share of female workers, total (RAIS) 0.29 0.2 0.0 1.0 (0.22) 961Weekly hours worked (RAIS) 43.51 44.0 30.0 44.0 (1.29) 961Weekly hours worked (WMS) 43.80 44.0 35.0 65.0 (2.47) 961
Worker educationShare of employees with university degree (WMS) 0.13 0.1 0.0 1.0 (0.13) 961Share of employees with university degree (RAIS) 0.13 0.1 0.0 1.0 (0.18) 961
Share of employees with high school degree (RAIS) 0.55 0.6 0.0 1.0 (0.21) 961Share of managers with university degree (WMS) 0.73 0.9 0.0 1.0 (0.33) 961Share of non-managers with university degree (WMS) 0.10 0.1 0.0 1.0 (0.13) 961
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 7 / 24
Introduction Data and summary stats Results Conclusion
Measuring “worker quality”: AKM decomposition
yit = α + xitβ + θi + ψJ(i,t) + εit .
yit – log monthly wage of worker i at time t
θi – worker effect (for managers and production workers, separately)
ψJ(i,t) – firm-specific contribution to pay
I Regress log wages on an experience profile (interacted with race and gender)
I Firm effect: how much more a worker gets paid moving from firm A to B.
I Worker effect: value of the worker’s portable skills as they across jobs.
Worker has some traits and skills that have some value in the labour market,and they carry that along with them.
I Lots of caveats here, but not the focus of the paper.
Distribution AKM table
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 8 / 24
Introduction Data and summary stats Results Conclusion
Firm data: World Management Survey
I WMS measures the level of adoption of structured management practices.
Main measure is the standardized average management score.
Includes 18 practices such as performance tracking and review, processdocumentation, target setting, people management.
I 1-hour interview with plant managers, responses scored on a 1-5 scale:
Score of 1: (“little/no formal management practices”)Score of 2 (“some informal management practices”)Score of 3: (“formal practices with some weaknesses”)Score of 4: (“established formal practices”)Score of 5: (“best practices, part of the culture of org”)
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 9 / 24
Introduction Data and summary stats Results Conclusion
Firm data: World Management Survey
I WMS measures the level of adoption of structured management practices.
Main measure is the standardized average management score.
Includes 18 practices such as performance tracking and review, processdocumentation, target setting, people management.
I 1-hour interview with plant managers, responses scored on a 1-5 scale:
Score of 1: (“little/no formal management practices”) — unstructuredScore of 2 (“some informal management practices”) — unstructuredScore of 3: (“formal practices with some weaknesses”)Score of 4: (“established formal practices”)Score of 5: (“best practices, part of the culture of org”)
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 9 / 24
Introduction Data and summary stats Results Conclusion
Firm data: World Management Survey
I WMS measures the level of adoption of structured management practices.
Main measure is the standardized average management score.
Includes 18 practices such as performance tracking and review, processdocumentation, target setting, people management.
I 1-hour interview with plant managers, responses scored on a 1-5 scale:
Score of 1: (“little/no formal management practices”) — unstructuredScore of 2 (“some informal management practices”) — unstructuredScore of 3: (“formal practices with some weaknesses”) — structuredScore of 4: (“established formal practices”) — structuredScore of 5: (“best practices, part of the culture of org”) — structured
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 9 / 24
Introduction Data and summary stats Results Conclusion
Distribution of Management Scores
Median0
.2.4
.6.8
Dens
ity
1 2 3 4 5Management score
Overall management
Note: Data from the WMS Brazil only. 2008 and 2013.
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 10 / 24
Introduction Data and summary stats Results Conclusion
Distribution of Management Scores
Median0
.2.4
.6.8
Dens
ity
1 2 3 4 5Management score
Overall managementPeople management
Note: Data from the WMS Brazil only. 2008 and 2013.
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 10 / 24
Introduction Data and summary stats Results Conclusion
PIA - Pesquisa Industrial Anual
I Industrial Survey of Manufacturing Establishments
I DesignI More than 30 Employees: CensusI 5-30 Employees: Random sample
I InformationI Longitudinal trackingI Balance sheetI Other economic variables
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 11 / 24
Introduction Data and summary stats Results Conclusion
1 Introduction
2 Data and summary statsWorkerManagementProductivity
3 Results
4 Conclusion
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 12 / 24
Introduction Data and summary stats Results Conclusion
Positive correlation btw productivity and better workers
9.5
1010
.511
11.5
12
Pro
duct
ivity
log
of v
alue
add
ed p
er e
mpl
oyee
-2 -1 0 1 2 3
Average employee fixed effect (standardized)
Overall PE Management PE Prod worker PE Tech worker PE
Note: productivity data from PIA, worker data from RAIS, management data from WMS. Raw data.N=108607. Both variables residualized by regressing the underlying variable on log employment.Binscatter by vingtiles of overall mean person effects.
table
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 13 / 24
Introduction Data and summary stats Results Conclusion
Positive correlation btw productivity and management
11.5
1212
.513
13.5
Pro
duct
ivity
log
of s
ales
per
em
ploy
ee
-2 -1 0 1 2
Management score (standardized): zmanagement
Note: productivity data from PIA, worker data from RAIS, management data from WMS. Collapsed by firm.N=720. Both variables residualized by regressing the underlying variable on log employment.
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 14 / 24
Introduction Data and summary stats Results Conclusion
Results: firms with structured management...
We find that more structured management is associated with:
1 Positive recruitment: hire a larger share of their new recruits from the topof the distribution of worker quality.
2 Worker matching and separation: fire less often and more selectively.
3 Retention practices: consistently retain a high share of the best workers.
4 Different practices are linked to selectivity across levels of the hierarchy.
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 15 / 24
Introduction Data and summary stats Results Conclusion
Results: firms with structured management...
We find that more structured management is associated with:
1 Positive recruitment: hire a larger share of their new recruits from the topof the distribution of worker quality.
2 Worker matching and separation: fire less often and more selectively.
3 Retention practices: consistently retain a high share of the best workers.
4 Different practices are linked to selectivity across levels of the hierarchy.
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 15 / 24
Introduction Data and summary stats Results Conclusion
... hire better workers
Managers
unstructuredmanagement
structuredmanagement
025
5075
100
Perc
entil
e ra
nk o
f hire
d m
anag
erw
ithin
firm
type
0 25 50 75 100Percentile rank of hired manager
in full distribution
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 16 / 24
Introduction Data and summary stats Results Conclusion
... hire better workers
Production workers
unstructuredmanagement
structuredmanagement
025
5075
100
Perc
entil
e ra
nk o
f hire
d pr
oduc
tion
wor
ker
with
in fi
rm ty
pe
0 25 50 75 100Percentile rank of hired production worker
in full distribution
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 16 / 24
Introduction Data and summary stats Results Conclusion
Results: firms with structured management...
We find that more structured management is associated with:
1 Positive recruitment: hire a larger share of their new recruits from the topof the distribution of worker quality.
2 Worker matching and separation: fire less often and more selectively.
3 Retention practices: consistently retain a high share of the best workers.
4 Different practices are linked to selectivity across levels of the hierarchy.
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 17 / 24
Introduction Data and summary stats Results Conclusion
... match and fire more selectively
.05
.1.1
5.2
Firi
ng r
ate
-1 0 1 2Worker quality
Unstructured management
Structured management
Managers
.05
.1.1
5.2
Firi
ng r
ate
-1 0 1 2Worker quality
Unstructured management
Structured management
Production workers
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 18 / 24
Introduction Data and summary stats Results Conclusion
Results: firms with structured management...
We find that more structured management is associated with:
1 Positive recruitment: hire a larger share of their new recruits from the topof the distribution of worker quality.
2 Worker matching and separation: fire less often and more selectively.
3 Retention practices: consistently retain a high share of the best workers.
4 Different practices are linked to selectivity across levels of the hierarchy.
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 19 / 24
Introduction Data and summary stats Results Conclusion
... retain better workers
Managers
low quality
high quality
low quality
high quality
0.5
1
2003 2005 2007 2009 2011 2013 2003 2005 2007 2009 2011 2013
Unstructured management Structured managementE
mpl
oym
ent s
tock
sha
re (
annu
al)
year
employment shares
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 20 / 24
Introduction Data and summary stats Results Conclusion
... retain better workers
Production workers
low quality
high quality
low quality
high quality
0.5
1
2003 2005 2007 2009 2011 2013 2003 2005 2007 2009 2011 2013
Unstructured management Structured managementE
mpl
oym
ent s
tock
sha
re (
annu
al)
year
employment shares
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 20 / 24
Introduction Data and summary stats Results Conclusion
Results
We find that more structured management is associated with:
1 Positive recruitment: hire a larger share of their new recruits from the topof the distribution of worker quality.
2 Worker matching and separation: fire less often and more selectively.
3 Retention practices: consistently retain a high share of the best workers.
4 Different practices are linked to selectivity across levels of the hierarchy.
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 21 / 24
Introduction Data and summary stats Results Conclusion
Selectivity driven by different types of practicesDependent variable: z-(production worker quality) z-(manager quality)
(1) (2) (3) (4) (5) (6) (7) (8)
Management indicesz-people 0.100*** 0.098** 0.086*** 0.007
(0.036) (0.042) (0.030) (0.035)z-operations 0.074** 0.004 -0.007 0.143*** 0.137*** 0.130***
(0.037) (0.043) (0.042) (0.033) (0.040) (0.040)
Individual practicesz-talent mindset 0.091** 0.059*
(0.036) (0.031)z-performance culture 0.016 0.008
(0.033) (0.032)z-talent capacity -0.027 -0.018
(0.033) (0.032)z-talent development 0.032 -0.013
(0.041) (0.035)z-value proposition 0.016 -0.015
(0.036) (0.037)z-retaining talent 0.017 -0.009
(0.038) (0.033)Firm controls Y Y Y Y Y Y Y YIndustry controls Y Y Y Y Y Y Y Y
# Observations 955 955 955 955 955 955 955 955# Firms 690 690 690 690 690 690 690 690R2 0.273 0.269 0.273 0.277 0.353 0.360 0.360 0.362
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 22 / 24
Introduction Data and summary stats Results Conclusion
Takeaways
Motivation:
I Building a productive workforce is key for productivity.
I Managers have a lot of discretion, but don’t always make the best decisions.
Can structured management practices help improve this?
We show that firms with more structured management tend to:
I Hire better (larger share from high quality worker distribution).
I Retain better (a higher share of high quality workers).
I Match better (lower turnover).
I Fire better (more selectively).
And also suggest that different practices matter for the selectivity across differentlevels of the hierarchy.
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 23 / 24
Introduction Data and summary stats Results Conclusion
Building a productive workforce:the role of structured management practices
Chris CornwellUniversity of Georgia
Ian SchmutteUniversity of Georgia
Daniela ScurMIT Sloan
ESCoE, NIESRApril 2019
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 24 / 24
APPENDIX
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 1 / 5
More productive firms use higher quality workers
Dependent variable: ln(sales) (1) (2) (3) (4) (5) (6)
Management scorez-management 0.213*** 0.168*** 0.088*** 0.065*** 0.064*** 0.059***
(0.039) (0.039) (0.02) (0.01) (0.01) (0.01)AKM quality measuresz-worker quality 0.247*** 0.076***
(0.039) (0.02)z-production worker quality 0.031** 0.028* 0.010
(0.02) (0.02) (0.02)z-manager quality 0.078*** 0.076*** 0.053***
(0.02) (0.02) (0.02)z-firm quality 0.098***
(0.02)Firm characteristicsShare workers with college degree 0.05 0.05
(0.10) (0.10)
Factor inputs Y Y Y YIndustry Y Y Y Y Y YOwnership Y Y Y Y Y Y# Observations 775 775 773 663 663 663# Firms 679 679 679 594 594 594R2 0.753 0.796 0.96 0.97 0.97 0.97
back
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 2 / 5
Share of the workforce captured by structured firms
By formal/informal structures
.3.4
.5.6
.7
2003 2008 20132003 2008 2013
Unstructured management Structured managementEm
ploy
men
t sha
re
Year
back
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 3 / 5
Share of the workforce captured by structured firms
By terciles
financialcrisis year
financialcrisis year
financialcrisis year.2
5.3
.35
.4
2003 2008 20132003 2008 20132003 2008 2013
Bottom tercile Middle tercile Top tercileEm
ploy
men
t sha
re
Year
back
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 3 / 5
Distribution of manager and non-manager fixed effects
0.5
11.
5D
ensi
ty
-2 0 2 4AKM worker fixed effect
Production workersManagers
back
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 4 / 5
Compensation practices, structured management, andemployment outcomes
Component Correlations
Variables Mean Std. Dev. Management ψ̂ ¯̂θManagement (z-score) 0 1 1.000
Firm effect: ψ̂ 0 1 0.439 1.000
Worker effect: ¯̂θ 0 1 0.392 0.603 1.000Firm size (employment) 583 787 0.325 0.219 0.267Share of employees with college degree 0.12 0.16 0.325 0.500 0.873Hire rate 0.19 0.14 -0.054 -0.193 -0.154Separation rate 0.22 0.13 -0.177 -0.251 -0.258Average tenure (months) 59.30 30.25 0.113 0.338 0.293Average hours worked (week) 43.56 1.32 -0.070 -0.273 -0.287Share of white workers 0.69 0.28 0.026 0.124 0.277Share of male workers 0.72 0.22 0.101 0.243 -0.013Share of unionized workers (non-managers) 0.56 0.40 -0.108 -0.123 -0.099
back
Cornwell, Schmutte, Scur Building a productive workforce RES Annual meeting, April 2019 5 / 5