1. The Effects of Management on Productivity:
Evidence from Mid-20th Century
Michela Giorcelli, UCLA and NBER
June 21, 2019
OECD Global Forum on Productivity
Sydney, Australia

2. Motivation
Large productivity spreads across firms (Syverson, 2004)
Persistent over time (Foster et al., 2008)
In developed and developing countries (Hsieh and Klenow, 2009)
Correlated with adoption of management practices (Bloom et al., 2007)
Limited causal effects (Bloom et al., 2013; Bruhn et al., 2018)
Implementation of managerial practices is firm choice
RCTs testing short-run impact
Offering in-plant managerial consulting
What can we learn from economic history?
Non-experimental setting with larger sample size
Plausibly exogenous variation
Possibility of evaluating heterogenous and long-term effects

3. THE LONG-TERM EFFECTS OF
MANAGEMENT AND
TECHNOLOGY TRANSFERS

4. The Marshall Plan Productivity Program
Transfer of US management and technology to Europe (1952-1958)
Management-training trips for European managers in US firms
Loans restricted to purchase technologically-advanced US machines
Data: 6,065 Italian firms eligible to participate in the program
Balance sheets from 5 years before to 15 after
Applications submitted for study trips and/or new machines
Identification strategy: Unexpected US budget cut and comparison
Firms that eventually participated
Firms initially eligible, excluded after the cut
That applied for the same US transfer before the cut

5. SME Manufacturing Firms Located in 5 Pilot Regions

6. Unexpected Budget Cut: 5 Experimental Provinces

7. Summary of the Results: Management Survival
Notes. Kaplan-Meier survivor function. Data are provided at firm level. The gray shaded
area corresponds to the three-year follow-up period after the US intervention.

8. Summary of the Results: Management Productivity
Notes. The dependent variables are logged TFPR, estimated with the Ackerberg et al.
(2006) method. Standard errors are block-bootstrapped.

9. Summary of the Results: Technology Productivity
Notes. The dependent variables are logged TFPR, estimated with the Ackerberg et al.
(2006) method. Standard errors are block-bootstrapped.

10. Mechanisms
Adoption of US management practices
More than 90% of firms within 3 years
Still in use 15 years later
Firm organization
Increase in number of plants and manager-to-worker ratio
More professionally-managed businesses
Financing and investment
Increase in bank credit, investment, ROA
No long-run effects for technology transfer

11. Open Questions
RCTs or business training programs provide consulting services
Management taught as bundle of practices
To top managers or owner-managers
Open questions
Effects of different management practices on firm performance?
Complementarity among management practices?
Similar effects if top or middle managers are trained?

12. NOT ALL MANAGEMENT IS
CREATED EQUAL:
EVIDENCE FROM THE
TRAINING WITHIN INDUSTRY
PROGRAM

13. TheTraining Within Industry Program
Management in-plant consulting service between 1940-1945
To US firms involved in war production
In three management areas, separately for top and middle managers
Factory operations (OP)
Human resources management (HR)
Inventory, sales, orders management (IO)
Organization of the program
TWI created “geographical” units called subdistricts
Voluntary basis, 10 application windows
Trained following application order, separately per window
Consulting provided by TWI instructors
Trained in one management area, for top or middle managers
Hired either full or part-time due to budget constraint
Could only train in one window

14. Identification: Assignment of Instructors to Subdistricts
ONLY based on number of applicant firms per subdistrict per window
Variation in full- vs part-time instructors and their training skills
Within window across subdistricts
Within subdistrict across windows
Determined firms eventually trained and type of training received
Comparing firms that applied on same date within county and sector
BUT assigned to different subdistricts
Received different trainings or eventually not treated
No correlation with pre-TWI firm characteristics
Data: 11,575 US firms that applied to TWI
Archival information on TWI training
Balance sheets and statement of profits and losses (1935-1955)
Survey data before and after the training

15. Summary Statistics in 1939 for 11,575 Applicant Firms
aaaaaaaaaaaaaaaaaaaaaaaaaaaa Mean St. Dev. Min. Max.
Plants 3.04 1.77 2 6
Employees 872.67 575.05 341 5,812
Current assets 25.32 8.90 17.89 37.65
Annual sales 23.84 10.13 15.68 43.56
Value added 8.58 5.75 5.67 14.81
Age 10.13 5.75 3 36
TFPR 3.12 0.78 1.87 4.09
Agriculture 0.05 0.06 0 1
Manufacturing 0.55 0.50 0 1
Transportation 0.28 0.28 0 1
Services 0.17 0.15 0 1
Observations 11,575 11,575 11,575 11,575
Notes. Summary statistics for 11,575 firms that applied to the TWI program. Data are pro-
vided at firm level. Current assets, Annual sales, and Value added are in 2019 USD; TFPR is
log total factor productivity revenue, estimated using the Ackerberg et al. (2006) method.

16. Training Received by 11,575 Applicant Firms
Notes. Type of training received by 11,575 firms that applied to the TWI program. NA is for
firms did not get any TWI intervention; Two is for firms received two-module trainings; All is
for firms received all three-module trainings; OP is for firms that received Factory Operation;
HR is for firms that received Human Resources; IO is for firms that received Inventory, Order
and Sales.

17. Empirical Specification
What is the effect of each separate training on firm performance?
outcomeit =
3
∑
λ=1
βλ(Treatmentλ
i · Postλ
it ) + Appl. Datei + δdst + it
where:
outcome: logged sales, TFPR, inventory, ROA
treatment: indicator for OP if λ = 1; HR if λ = 2; IO if λ = 3
Postλ
it: indicator for years after receiving TWI training λ
Appl. Datei : firm application date
δdst: district-sector-year fixed effects
Treatment: firms that got only one intervention
Comparison: firms never treated
Standard errors clustered at subdistrict level

18. Identification Assumptions
βλ estimate causal effect of TWI training λ if trend in performance of
treated and comparison firms would have been the same without the
TWI
Supportive evidence:
1. Firm characteristics do not predict instructors assignments
2. Application timing does not predict probability of getting treated
3. Firm characteristics statistically equivalent in 1939
4. On the same trend in 1935-1939

19. Effects of OP: 1.5-2.5% Increase in Performance per Year
Sales TFPR Inventory ROA
OP*post 0.025*** 0.022*** -0.002 0.015***
(0.006) (0.005) (0.005) (0.005)
HR*post 0.054*** 0.045*** 0.004 0.038***
(0.005) (0.007) (0.006) (0.006)
IO*post 0.032*** 0.037*** -0.041*** 0.025***
(0.004) (0.006) (0.006) (0.005)
District-sector-year FE Yes Yes Yes Yes
Test OP=HR 78.91 88.72 1.96 57.89
Test HR=IO 61.23 92.34 45.67 66.78
Test OP=IO 55.46 78.34 44.53 88.45
Observations 145,480 145,480 145,480 145,480
Notes. TFPR is log total factor productivity revenue (Ackerberg et al, 2016); Sales
and Inventory are expressed in million 2019 USD. ROA is the ratio between profits
and assets. Standard errors are clustered at the subdistrict level.

20. Effects of HR: 3.8-5.4% Increase in Performance per Year
Sales TFPR Inventory ROA
OP*post 0.025*** 0.022*** -0.002 0.015***
(0.006) (0.005) (0.005) (0.005)
HR*post 0.054*** 0.045*** 0.004 0.038***
(0.005) (0.007) (0.006) (0.006)
IO*post 0.032*** 0.037*** -0.041*** 0.025***
(0.004) (0.006) (0.006) (0.005)
District-sector-year FE Yes Yes Yes Yes
Test OP=HR 78.91 88.72 1.96 57.89
Test HR=IO 61.23 92.34 45.67 66.78
Test OP=IO 55.46 78.34 44.53 88.45
Observations 145,480 145,480 145,480 145,480
Notes. TFPR is log total factor productivity revenue (Ackerberg et al, 2016); Sales
and Inventory are expressed in million 2019 USD. ROA is the ratio between profits
and assets. Standard errors are clustered at the subdistrict level.

21. Effects of IO: 2.5-3.7% Increase in Performance per Year
Sales TFPR Inventory ROA
OP*post 0.025*** 0.022*** -0.002 0.015***
(0.006) (0.005) (0.005) (0.005)
HR*post 0.054*** 0.045*** 0.004 0.038***
(0.005) (0.007) (0.006) (0.006)
IO*post 0.032*** 0.037*** -0.041*** 0.025***
(0.004) (0.006) (0.006) (0.005)
District-sector-year FE Yes Yes Yes Yes
Test OP=HR 78.91 88.72 1.96 57.89
Test HR=IO 61.23 92.34 45.67 66.78
Test OP=IO 55.46 78.34 44.53 88.45
Observations 145,480 145,480 145,480 145,480
Notes. TFPR is log total factor productivity revenue (Ackerberg et al, 2016); Sales
and Inventory are expressed in million 2019 USD. ROA is the ratio between profits
and assets. Standard errors are clustered at the subdistrict level.

22. No Effects on War-Related Outcomes
Government Number Value Post-WWII
Sales War Contr. War Contr. Refunds
OP*post 0.003 -0.002 0.004 -0.003
(0.004) (0.005) (0.007) (0.004)
HR*post 0.002 0.004 0.007 0.004
(0.004) (0.007) (0.009) (0.007)
IO*post -0.002 -0.003 -0.002 -0.001
(0.005) (0.005) (0.004) (0.002)
Dsy FE Yes Yes Yes Yes
Obs. 36,370 36,370 36,370 29,096
Notes. Government sales (m USD), expressed in million 2019 USD, are the sales
made directly to the government; Government TFPR is the logarithm of total factor
productivity revenue, estimated using the Ackerberg et al. (2006), using government
revenues only; N War Contracts and Value War Contracts are the number and value
of war supply contracts granted to a firm. Post-War Refunds are subsidies given by
the government to war contractors to switch from military to war production. Stan-
dard errors are clustered at the subdistrict level.

23. Changes in OP Managerial Practices With OP Training
Managerial Practicesaaaaaaaaaaaa OP*post HR*post IO*post
(1) Machine Repairs -0.248*** 0.005 -0.002
(0.059) (0.006) (0.004)
(2) Worker’s Injuries -0.332*** -0.003 0.004
(0.065) (0.004) (0.005)
(3) Causes of Breakdown 0.751*** -0.002 0.003
(0.212) (0.005) (0.004)
Observations 27,506 27,506 27,506
Notes. Each row represents a separate regressions whose dependent variables are each
of the 11 management practices. Data are provided at the plant level. Standard errors
are clustered at the subdistrict level.

24. Changes in HR Managerial Practices With HR Training
Managerial Practicesaaaaaaaaaaaa OP*post HR*post IO*post
(4) Job Description Managers 0.003 0.922*** -0.002
(0.005) (0.234) (0.003)
(5) Job Description Workers -0.005 0.943*** 0.003
(0.007) (0.321) (0.005)
(6) Training for Workers 0.007 0.891*** -0.004
(0.006) (0.289) (0.006)
(7) Introduction of Bonus 0.002 0.873*** 0.005
(0.003) (0.342) (0.006)
(8) Suggestions from Workers 0.003 0.556*** 0.004
(0.005) (0.342) (0.005)
Observations 27,506 27,506 27,506
Notes. Each row represents a separate regressions whose dependent variables are each
of the 11 management practices. Data are provided at the plant level. Standard errors
are clustered at the subdistrict level.

25. Changes in IO Managerial Practices With IO Training
Managerial Practicesaaaaaaaaaaaa OP*post HR*post IO*post
(9) Unused Input -0.005 0.004 -0.678***
(0.006) (0.007) (0.003)
(10) Production Planning 0.006 0.006 0.893***
(0.009) (0.005) (0.003)
(11) Marketing -0.004 -0.004 0.851***
(0.009) (0.005) (0.246)
Observations 27,506 27,506 27,506
Notes. Each row represents a separate regressions whose dependent variables are each
of the 11 management practices. Data are provided at the plant level. Standard errors
are clustered at the subdistrict level.

26. Complementarity Effects of HR
aaaaaaaaaaaaaaaaaaaa Sales TFPR Inventory ROA
HR after OP*post 0.065*** 0.058*** -0.005 0.052***
(0.007) (0.006) (0.006) (0.005)
HR after IO*post 0.074*** 0.065*** 0.004 0.068***
(0.005) (0.007) (0.006) (0.006)
HR*post 0.055*** 0.047*** -0.003 0.040***
(0.008) (0.008) (0.005) (0.007)
D-s-y FE Yes Yes Yes Yes
Test HR after OP=HR 43.56 62.84 2.77 54.31
Test HR after IO=HR 62.25 81.48 3.21 50.71
Observations 198,720 198,720 198,720 198,720
Notes. TFPR is log total factor productivity revenue (Ackerberg et al, 2016); Sales
and Inventory are expressed in million 2019 USD. ROA is the ratio between profits
and assets. Standard errors are clustered at the subdistrict level.

27. Complementary Effects ONLY with HR
aaaaaaaaaaaaaaaaaaaa Sales TFPR Inventory ROA
OP after HR*post 0.038*** 0.033*** 0.002 0.022***
(0.005) (0.007) (0.004) (0.004)
OP after IO*post 0.026*** 0.019*** -0.004 0.014***
(0.005) (0.004) (0.005) (0.003)
OP *post 0.026*** 0.021*** 0.001 0.013***
(0.007) (0.004) (0.002) (0.003)
D-s-y FE Yes Yes Yes Yes
Test OP after HR=OP 48.68 53.01 2.69 61.25
Test OP after IO=OP 2.48 1.43 1.27 2.49
Observations 198,720 198,720 198,720 198,720
Notes. TFPR is log total factor productivity revenue (Ackerberg et al, 2016); Sales
and Inventory are expressed in million 2019 USD. ROA is the ratio between profits
and assets. Standard errors are clustered at the subdistrict level.

28. Complementary Effects ONLY with HR
aaaaaaaaaaaaaaaaaaaa Sales TFPR Inventory ROA
IO after HR*post 0.049*** 0.041*** -0.055*** 0.053***
(0.005) (0.007) (0.010) (0.006)
IO after OP*post 0.031*** 0.028*** -0.041*** 0.026***
(0.006) (0.005) (0.005) (0.006)
IO*post 0.030*** 0.035*** -0.044*** 0.024***
(0.006) (0.007) (0.008) (0.004)
D-s-y FE Yes Yes Yes Yes
Test IO after HR=IO 77.41 53.96 60.40 70.08
Test IO after OP=IO 2.77 1.98 1.29 2.64
Observations 198,720 198,720 198,720 198,720
Notes. TFPR is log total factor productivity revenue (Ackerberg et al, 2016); Sales
and Inventory are expressed in million 2019 USD. ROA is the ratio between profits
and assets. Standard errors are clustered at the subdistrict level.

29. Should We Train Top or Middle Managers?
Factory operations: Similar effects
Basic tasks implemented by low-skilled workers
Human resources: Larger effects for middle managers
More interactions with non-managerial staff
Survey data: workers suggestions
74% implemented by middle managers
37% implemented by top managers
Inventory, sales and orders: Larger effects for top managers
Involved higher-level business choices
Survey data: production planning
79% implemented by middle managers
94% implemented by top managers

30. OP: Similar Effects
aaaaaaaaaaaaaaaaa Sales TFPR Inventory ROA
TopOP*post 0.022*** 0.020*** -0.001 0.014***
(0.005) (0.004) (0.002) (0.003)
MiddleOP*post 0.026*** 0.023*** 0.003 0.016***
(0.004) (0.006) (0.003) (0.004)
TopHR*post 0.035*** 0.029*** -0.001 0.025***
(0.005) (0.006) (0.002) (0.007)
MiddleHR*post 0.067*** 0.056*** 0.004 0.045***
(0.010) (0.008) (0.005) (0.009)
TopIO*post 0.040*** 0.043*** -0.062*** 0.033***
(0.004) (0.006) (0.006) (0.005)
MiddleIO*post 0.020*** 0.027*** -0.031*** 0.024***
(0.005) (0.005) (0.006) (0.009)
Observations 145,480 145,480 145,480 145,480
Notes. TFPR is log total factor productivity revenue (Ackerberg et al, 2016); Sales and In-
ventory are expressed in million 2019 USD. ROA is the ratio between profits and assets. Stan-
dard errors are clustered at the subdistrict level.

31. HR: Larger Effects for Middle Managers
aaaaaaaaaaaaaaaaa Sales TFPR Inventory ROA
TopOP*post 0.022*** 0.020*** -0.001 0.014***
(0.005) (0.004) (0.002) (0.003)
MiddleOP*post 0.026*** 0.023*** 0.003 0.016***
(0.004) (0.006) (0.003) (0.004)
TopHR*post 0.035*** 0.029*** -0.001 0.025***
(0.005) (0.006) (0.002) (0.007)
MiddleHR*post 0.067*** 0.056*** 0.004 0.045***
(0.010) (0.008) (0.005) (0.009)
TopIO*post 0.040*** 0.043*** -0.062*** 0.033***
(0.004) (0.006) (0.006) (0.005)
MiddleIO*post 0.020*** 0.027*** -0.031*** 0.024***
(0.005) (0.005) (0.006) (0.009)
Observations 145,480 145,480 145,480 145,480
Notes. TFPR is log total factor productivity revenue (Ackerberg et al, 2016); Sales and In-
ventory are expressed in million 2019 USD. ROA is the ratio between profits and assets. Stan-
dard errors are clustered at the subdistrict level.

32. IO: Larger Effects for Top Managers
aaaaaaaaaaaaaaaaa Sales TFPR Inventory ROA
TopOP*post 0.022*** 0.020*** -0.001 0.014***
(0.005) (0.004) (0.002) (0.003)
MiddleOP*post 0.026*** 0.023*** 0.003 0.016***
(0.004) (0.006) (0.003) (0.004)
TopHR*post 0.035*** 0.029*** -0.001 0.025***
(0.005) (0.006) (0.002) (0.007)
MiddleHR*post 0.067*** 0.056*** 0.004 0.045***
(0.010) (0.008) (0.005) (0.009)
TopIO*post 0.040*** 0.043*** -0.062*** 0.033***
(0.004) (0.006) (0.006) (0.005)
MiddleIO*post 0.020*** 0.027*** -0.031*** 0.024***
(0.005) (0.005) (0.006) (0.009)
Observations 145,480 145,480 145,480 145,480
Notes. TFPR is log total factor productivity revenue (Ackerberg et al, 2016); Sales and In-
ventory are expressed in million 2019 USD. ROA is the ratio between profits and assets. Stan-
dard errors are clustered at the subdistrict level.

33. Interpretation of the Results
Factory operations: Similar effects
Basic tasks implemented by low-skilled workers
Human resources: Larger effects for middle managers
More interactions with non-managerial staff
Survey data: workers suggestions
74% implemented by middle managers
37% implemented by top managers
Inventory, sales and orders: Larger effects for top managers
Involved higher-level business choices
Survey data: production planning
79% implemented by middle managers
94% implemented by top managers

34. Conclusions and Discussion
Positive effects of managerial practices
Persistent in the long-run
Different magnitude and complementarity effects
Heterogenous effects if top or middle managers are trained
Implications for public policies
Business training programs largely used today
Most successful combination of managerial practices
Type of managers to target
External validity
Management practices taught similar across time
Informative for multi-plant organizations
Managerial practices spread remain fairly large