motivation: what is management?

MANAGEMENT AS A TECHNOLOGY?

Nick Bloom (Stanford), Raffaella Sadun (HBS) and

John Van Reenen (LSE)

Motivation: What is Management?

• Evidence of massive national and plant spread in TFP:e.g. Hall and Jones (1999) Syverson (2004)

• Can management explain any of this: 1%, 10% or 100%?

• Two main theories take opposite positions:– Management as Technology (MAT): management key

driver of performance gaps (Walker, 1887, Marshal 1887)

– Management as Design (MAD): management optimized so less important (Woodward, 1958, core Org-Econ)

Ohio, USA Maharashtra, India

MAT v MAD – are management difference optimal?

Summary of the paper

1. Management data from ≈ 15,000 firms in 30 countries– US highest (unweighted) average management score– US lead ≈50% larger if size weight (more reallocation)

2. Are these variations technology (MAT) or optimal (MAD)? Develop three predictions on: performance, competition and reallocation, and find best fit for MAT

3. Given MAT, then estimate management can account for possibly ≈ 25% of plant and country spread in TFP

Management Models

Testing the models

Management Data

Management and cross-country and firm TFP

1) Developing management questions

•Scorecard for 18 monitoring and incentives practices in ≈45 minute phone interview of manufacturing plant managers

2) Getting firms to participate in the interview

•Introduced as “Lean-manufacturing” interview, no financials

•Official Endorsement: Bundesbank, RBI, PBC, World Bank etc.

3) Obtaining unbiased comparable responses, “Double-blind”•Interviewers do not know the company’s performance

•Managers are not informed (in advance) they are scored

Survey methodology (following Bloom & Van Reenen (2007))

Score (1): Measures tracked do not indicate directly if overall business objectives are being met. Certain processes aren’t tracked at all

(3): Most key performance indicators are tracked formally. Tracking is overseen by senior management

(5): Performance is continuously tracked and communicated, both formally and informally, to all staff using a range of visual management tools

Example monitoring question, scored based on a number of questions starting with “How is performance tracked?”

Score (1) People are promoted primarily upon the basis of tenure, irrespective of performance (ability & effort)

(3) People are promoted primarily upon the basis of performance

(5) We actively identify, develop and promote our top performers

Example incentives question, scored based on questions starting with “How does the promotion system work?”

Survey randomly drawn firms from the population of larger (50 to 5,000 employee) manufacturers across countries (including public and private firms)

Locations of plants surveyed 2004-2008

Wide spread of management: US, Japan & Germany leading, and Africa, South America and India trailing

Data includes 2013 survey wave as of 9/20/2013. Africa data not yet included in the paper

N=80N=87N=122

N=74N=50

N=120N=127

N=840N=558N=1111

N=755N=581N=269

N=160N=307

N=136N=150N=306

N=515N=364

N=454N=313N=632N=1208

N=412N=403

N=658N=176

N=1289

2 2.5 3 3.5Average Management Scores, Manufacturing

TanzaniaGhana

EthiopiaNicaragua

ZambiaKenya

ColombiaIndia

ArgentinaBrazilChinaChile

GreeceRepublic of Ireland

PortugalNorthern Ireland

New ZealandSingapore

MexicoPoland

AustraliaItaly

FranceGreat Britain

CanadaSweden

GermanyJapan

United States

Note: Firms between 50 and 5000 employees, Raw data

Africa

Asia

Australasia

Europe

Latin America

North America

Some quotes illustrate the African management approach

Interviewer “What kind of Key Performance Indicators do you use for performance tracking?”

Manager: “Performance tracking? That is the first I hear of this. Why should we spend money to hire someone to track our performance? It is a waste of money!”

Interviewer “How do you identify production problems?”

Production Manager: “With my own eyes. It is very easy”

Average management scores across countries are strongly correlated with GDP


Ethiopia Ghana

Kenya

Tanzania

Zambia

Australia

New Zealand

China

India

Japan

Singapore

France

Germany

Greece

Italy

Poland

PortugalRepublic of Ireland

Sweden

United Kingdom

ArgentinaBrazilChile

Colombia

Mexico

Nicaragua

Canada

United States

22

.53

3.5

Ave

rag

e m

anag

eme

nt p

ract

ice

s

7 8 9 10 11Log of 10-yr average GDP based on PPP per capita GDP(Current int'l $ - Billions)

Africa

Australasia

Asia

Europe

Latin America

North America

management x log of GDP PPP per capita

Note: April 2013, World Economic Outlook (IMF) indicator

Like TFP management also varies within countries


0.5

10

.51

0.5

10

.51

0.5

1

1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 United States 2 Japan 3 Germany 4 Sweden 5 Canada 6 Great Britain

7 France 8 Italy 9 Australia 10 Poland 11 Mexico 12 Singapore

13 New Zealand 14 Northern Ireland 15 Portugal 16 Republic of Ireland 17 Greece 18 Chile

19 China 20 Brazil 21 Argentina 22 India 23 Colombia 24 Kenya

25 Zambia 26 Nicaragua 27 Ethiopia 28 Ghana 29 Tanzania

Ke

rne

l De

nsi

ty E

stim

atio

n

Firm Average Management ScoreGraphs by rank_cty

Management Models

Testing the models

Management Data


Two perspectives on these management differences in economics

• Management as a Technology (MAT)– Management part of firm’s TFP, often said to explain

a firm’s TFP fixed effect (Mundlak, 1961)– Think of improvements as “process innovations”,

hence the sense that management is a technology

“It is on account of the wide range [of ability] among the employers of labor that we have the phenomenon in every community and in every trade some employers realizing no profits at all, while others are making fair profits; others, again, large profits; others, still, colossal profits.”

Management as a technology: Francis Walker

Francis Walker (QJE, April 1887)

Walker ran the 1870 US Census, and was the founding president of the AEA and president of MIT

“I am very nearly in agreement with General Walker’s Theory of profits….the earnings of management of a manufacturer represents the value of the addition which his work makes to the total produce of capital and industry....”

Alfred Marshall (QJE, July 1887)

Management as a technology: Alfred Marshall

• Management as a Technology (MAT)– Management part of firm’s TFP, often exampled as

explaining firm’s TFP fixed effects (Mundlak, 1961)– Think of improvements as “process innovations”, hence

the sense that management is a technology

• Management as Design (MAD)– Management is optimally chosen– Popular in Organizational Economics (Gibbons and

Roberts HOE, 2013), and other fields like Strategy and Management (Contingency theory, Woodward 1958)

Two perspectives on these management differences in economics

So we define 2 highly stylized management models

Production Function: Y=AKαLβG(M), M = management

Flavor 1: Management as Technology (MAT)

G(M)=Mγ ; pay for M which depreciates (like R&D)

Flavor 2: Management as Design (MAD)

G(M) = 1/(1+|M*-M|); M* = “optimal” choice, M is free

•Otherwise common set of assumptions–Changing M & K involves adjustment costs (L flexible)–τ % of sales lost to distortions (bribes, regulations etc)–Monopolistic competition (Isoelastic demand)–Sunk entry cost (E) & fixed per period operating cost (F)

We simulate these two management models

1. Entrants pay a sunk cost E for a draw on (A,M,τ), with rate of entry determined by the free entry condition

2. All firms get a shock to TFP: At=ρAt-1 + εt

3. Pay fixed operating cost F per period (or exit)

4. Invest in M & K, plus choose optimal labor

Result 1: Performance (size, TFP, profits) increase in management with MAT but not MAD

0.2

.4.6

.81

Sal

es

1 2 3 4 5Management

0.2

.4.6

.81

Sal

es

1 2 3 4 5Management

MAT: Y=AKαLβMγ MAD: Y=AKαLβ/(1+|M-M*|)

Notes: Results from simulating 2,500 firms per year in the steady state taking the last 5 years of data. Simulations are identical except that the production functions differ as outlined above. Plots normalized log(management) in the simulation data onto a 1 to 5 scale and log(sales) onto a 0 to 1 scale. Lowess plots shown with Stata defaults (bandwidth of 0.8 and tricube weighting).

Result 2: Average Management increases with competition (elasticity) with MAT but not MAD

Notes: Results from simulating 2,500 firms per year in the steady state taking the last 5 years of data for each level of elasticity. Simulations are identical for MAT and MAD except that the production functions differ as outlined above. Plots normalized ln(management) in the simulation data onto a 1 to 5 scale.

11.

52

2.5

33.

54

mea

n of

ilog

man

agem

ent

2 4 6 8 10

11.

52

2.5

33.

54

mea

n of

ilog

man

agem

ent

8 2 4 10 6

Comp increasing Comp increasing

MAT: Y=AKαLβMγ MAD: Y=AKαLβ/(1+|M-M*|)

Result 3: Sales weighted management is higher in less distorted economies with MAT but not MAD

3.8

3.9

44.

14.

2m

ean

of il

ogm

anag

emen

t

2 5 10 15 20

Notes: Plots log(management) scores weighted by firm sales. Results from simulating 2,500 firms per year in the steady state taking the last 5 years of data for each level of elasticity. Simulations are identical for MAT and MAD except that the production functions differ as outlined above. ln(management) in the simulation data is normalized onto a 1 to 5 scale.

3.8

3.9

44.

14.

2m

ean

of il

ogm

anag

emen

t

2 5 10 15 20

Distortions increasing(% sales lost)

Distortions increasing(% of sales lost)

MAT: Y=AKαLβMγ MAD: Y=AKαLβ/(1+|M-M*|) US

India

Very stylized models with obvious extensions

• Dynamics: maybe management also changes adjustment costs, information (forecasting) and factor prices

• Multi-factor: currently 1-dimensional management, but design view model could also be about sub-components

• Spillovers: Technology is (partially) non-rival so we should see learning, information spillovers, clustering etc.

• Governance/ownership issues: family firms, FDI, etc.

• Co-ordination: e.g. Gibbons & Henderson (2012)

Testing the models• Performance• Competition• Reallocation

Management Models

Measuring Data


Dependentvariable

Ln(sales) Ln(sales) Ln(employ-ment)

Profit rateROCE

5yr Salesgrowth

Exit Exit

OLSFixed

EffectsOLS OLS OLS OLS OLS

Firm sample All 2+ surveys All All Quoted All All

Management 0.150*** 0.033** 0.338*** 1.202*** 0.039*** -0.006*** -0.002

Ln(emp) 0.645*** 0.374***

Ln(capital) 0.307*** 0.237***

Competition 0.094**

Management*competition

-0.096**

Obs 8,314 6,364 15,608 9,163 8,365 7,532 7,532

TABLE 2: Performance is robustly correlated with management (consistent with MAT)

M, Management Index is average of all 18 questions (sd=1). Other controls include % employees with college degree, average hours worked, firm age, industry, country & time dummies & noise (e.g. interviewer dummies). Standard errors clustered by firm.

Dependent variable Growth in firm sales

Shock (Industry Sales) -0.033** -0.035**(0.014) (0.014)

Management*SHOCK 0.027**(0.011)

Shock (Industry Trade) -0.051*** -0.052***(0.014) (0.014)

Management*SHOCK 0.018*(0.010)

Management 0.002 -0.014 0.001 -0.008(0.006) (0.010) (0.006) (0.009)

Firms 1,567 1,567 1,599 1,599Observations 1,653 1,653 1,685 1,685

TABLE 3: Performance: use the Great Recession as a natural experiment and find again well managed firms perform better

Notes: SHOCK is a binary indicator for a fall in exports or fall in sales in the SIC3 by country cell from 2007 to 2009. All columns include controls for skills, firm and plant size, noise, country and industry dummies. Management from 2004-2006

Dependent Variable Ln(Sales) Ln(Sales) Ln(Sales) ROCE Growth Exit

CellSIC3

×CountryCountry SIC3

SIC3×Cty

SIC3×Cty

SIC3×Cty

Management 0.236*** 0.219*** 0.226*** 1.681*** 0.038 -0.018**

*I(M below cell ave) (0.032) (0.034) (0.039) (0.528) (0.028) (0.005)

Management 0.226*** 0.219*** 0.223*** 1.679*** 0.044* -0.014**

*I(M above cell ave) (0.027) (0.030) (0.030) (0.436) (0.024) (0.004)

F test of symmetry 0.230 0.984 0.785 0.990 0.388 0.060above v below cell aveCell Clusters 796 18 177 900 921 1,137

Observations 8,003 8,314 8,292 8,793 8,007 6,607

Notes: Regressions includes controls for country, SIC4 & year dummies, firm-age, average hours, % with degrees, noise controls etc. SE clustered by firm. Productivity columns from regression of ln(sales) as dependent variable with controls for ln(labor) and ln(capital). Only cells with 2+ observations used.

Table 4: Contingent performance: not consistent with MAD (best to be at optimal industry/country point)


Management Models

Measuring Data

Management and cross-country TFP

Competition proxies

Management(estimated in levels)

Management(estimated in differences)

Import penetration

0.805***(0.236)

0.608***(0.230)

1- Lerner Index1 17.53*

(3.85)20.68**(6.65)

# of competitors

0.121***(0.023)

0.120**

(0.052)

Balanced panel

No No No Yes Yes Yes

Obs 2,657 2,819 2,789 412 429 432

TABLE 5: Competition improves management (consistent with MAT)

Notes: Includes SIC-3 industry, country, firm-size, public and interview noise (interviewer, time, date & manager characteristic) controls. Col 1,2, 4 & 5 clustered by industry*country, cols 3 & 6 by firm


Management Models

Measuring Data


Dependent Variable Employees Employees Sales GrowthManagement (MNG,US base) 201.9*** 359.7*** 0.092***MNG*Argentina -270.9** -0.134***MNG*Australia -259.3* -0.145**MNG*Brazil -211.7* -0.101***MNG*Canada -169.3 -0.131**MNG*Chile -92.6 -0.150MNG*China -84.9 -0.060MNG*France -489.5** -0.085*MNG*Germany -9.0 -0.080*MNG*Greece -355.9*** -0.089**MNG*India -145.4 -0.066MNG*Ireland -258.8** -0.085MNG*Italy -283.1*** -0.092**MNG*Mexico -250.1* -0.075*MNG*New Zealand -375.7* 0.718***MNG*Japan -297.3** -0.099**MNG*Poland -308.1*** -0.058MNG*Portugal -308.9*** -0.109**MNG*Sweden -228.7* -0.068MNG*UK -125.1 -0.054Observations 5,842 3,858 2,756

Table 6: More reallocation to better managed firms in the US where markets generally less distorted (consistent with MAT)

Notes: Includes year, country, three digit industry dummies, # management questions missing, firm age, skills and noise controls.

Dependent Variable: EmploymentManagement (MNG) 231.46*** 356.73*** 110.97* (37.12) (55.89) (66.30)Management*Employment Protection -1.43** (World Bank Country Index) (0.69)

Management*Trade Costs -0.18*** (World Bank Country Index) (0.69) (0.05) Tariffs -4.96(country x country) (4.12)Management*Tariff -8.25**

(3.35)Observations 5,760 5,017 1,559

Notes: OLS, clustered by firm; dependent variable is firm employment; Domestic firms only. Controls for firm age, skills, noise, SIC3, country dummies, EPL(WB) is “difficulty of hiring” from World Bank (1=low, 100=high). Trade cost from World Bank (0=low,6=high). Last columns tariffs are in deviations from industry and country specific mean, from Feenstra and Romalis (2012).

Table 7: Reallocation also stronger in countries with lower labor and trade regulations (consistent with MAT)

Testing the model• Performance• Competition• Reallocation• Extensions: skills

Management Models

Measuring Data


Management and Education: use UNESCO World Higher Education Database university locations (N=9,081)

Dependent

Variable:

Manage

ment

% firm employees with degree

Manage

ment

Manage

ment

OLS OLS OLS IV

Drive time to nearest -0.049*** -1.534***

university (0.019) (0.423)

% employees with 0.789*** 3.190***

degree in the firm (0.082) (1.113)

Observations 6,406 6,406 6,406 6,406

Notes: Clustered by 313 regions. In final column proportion skilled is instrumented with distance to university. Include industry, regional and full set of firm and noise controls.

Distance to the nearest university seems to matter for firm skills and management

Testing the model• Performance• Competition• Reallocation• Extensions: skills

Management Models

Measuring Data

Management and cross firm and country TFP

Following MAT we can estimate rough contribution of management to country TFP spread

1. Estimate country differences in size weighted management

2. Impute impact of this on differences in TFP

Requires many assumptions, so only magnitude calculation

Ma

na

gem

en

t ga

p w

ith th

e U

S

Notes: Total weighted mean management deficit with the US is the number on top of bar. This is decomposed into (i) reallocation effect (blue bar) and (ii) unweighted average management scores (red bar) . Domestic firms, scores corrected for sampling bias

First calculate the employment weighted difference in management (from the US as baseline)

-.49

-1.65

-.27

-1.2

-.22

-1.19

-.26

-1.18

-.12

-1.02

-.26

-.98

-.29

-.83

-.18

-.81

-.19

-.74

-.11

-.49

-.08

-.36

-.18

-.35

-.14

-.25

00

-2-1

.5-1

-.5

0

grptarcnbrfritpogbcagejpswus

mean of rel_OP mean of rel_zmanReallocation Within Firm

Ma

na

gem

en

t ga

p w

ith th

e U

S

-.49

-1.65

-.27

-1.2

-.22

-1.19

-.26

-1.18

-.12

-1.02

-.26

-.98

-.29

-.83

-.18

-.81

-.19

-.74

-.11

-.49

-.08

-.36

-.18

-.35

-.14

-.25

00

-2-1

.5-1

-.5

0

grptarcnbrfritpogbcagejpswus

mean of rel_OP mean of rel_zmanReallocation Within Firm

30% of US-Greece management gap due to better US reallocation

Notes: Total weighted mean management deficit with the US is the number on top of bar. This is decomposed into (i) reallocation effect (blue bar) and (ii) unweighted average management scores (red bar) . Domestic firms, scores corrected for sampling bias

First calculate the employment weighted difference in management (from the US as baseline)

Country

Share-WeightedAverage Management

Deficit with US

TFP GAP with US

Proportion of TFP gap due to

ManagementUS 0Sweden -0.25 32.2 7.8%Japan -0.35 33.6 10.4%Canada -0.50 22.3 22.4%Great Britain -0.74 20.3 36.5%Italy -0.81 17.2 47.7%France -0.82 25.3 38.7%Brazil -0.98 59.6 16.9%China -1.01 78.3 14.9%Argentina -1.17 57.3 20.6%Portugal -1.18 24.9 48.2%Greece -1.65 51.0 32.4%Unweighted av. 25%

Assume one sd increase in management increases TFP by 10%. Regressions suggest about 5% to 15% depending on specification. TFP data from Jones and Romer (2010).

Second, estimate impact of management on TFP using result from micro regressions and field experiments result: ↑1 SD management ≈ ↑ 10% TFP

• Typical 4 digit industry in US has TFP ratio of 2:1 for 90-10 (Syversson, 2004)

• 90-10 for management is about 2.6 standard deviations

• Using a causal effect of 10% this implies management causes ≈ ¼ of TFP dispersion

Interestingly, also get similar 25% share for cross firm contribution of management to TFP spread

CONCLUSIONS

• Large spreads in size weighted management across firms and countries, with about 1/3 of US lead due to reallocation

• Micro data consistent with a model in which this variation in management is technology, with better & worse practices

• Estimate variations in management accounts for ≈25% of plant and country variation in TFP

MY FAVOURITE QUOTES:

Interviewer: “How many production sites do you have abroad?

Manager in Indiana, US: “Well…we have one in Texas…”

Americans on geography

Production Manager: “We’re owned by the Mafia”

Interviewer: “I think that’s the “Other” category……..although I guess I could put you down as an “Italian multinational” ?”

The difficulties of defining ownership in Europe

Interviewer : “Do staff sometimes end up doing the wrong sort of work for their skills?

NHS Manager: “You mean like doctors doing nurses jobs, and nurses doing porter jobs? Yeah, all the time. Last week, we had to get the healthier patients to push around the beds for the sicker patients”

Don’t get sick in Britian


Don’t do Business in Indian hospitals

Interviewer: “Is this hospital for profit or not for profit”

Hospital Manager: “Oh no, this hospital is only for loss making”

Interviewer : “Do you offer acute care?”

Switchboard: “Yes ma’am we do”

Don’t get sick in India


Interviewer : “Do you have an orthopeadic department?”

Switchboard: “Yes ma’am we do”

Interviewer : “What about a cardiology department?”

Switchboard: “Yes ma’am”

Interviewer : “Great – can you connect me to the ortho department”

Switchboard?: “Sorry ma’am – I’m a patient here”

[( )( )]

i ii

i i i ii

M M Y

M M Y Y M

OP M

“OLLEY PAKES” (OP) DECOMPOSITION OF WEIGHTED AVERAGE MANAGEMENT SCORE (M) IN GIVEN COUNTRY

Overall management z-score of firm i Size of firm i

Covariance(Olley-Pakes, 1996, reallocation term)

Unweighted mean of management score

( ) ( )k US k US k USM M OP OP M M

DECOMPOSING THE RELATIVE MANAGERIAL DEFICIT BETWEEN COUNTRY j AND THE US ECONOMY

Difference in reallocation(between firm)

Difference in unweightedMeans (within firm)

Difference in aggregateshare-weighted Management scores

Country

Share-WeightedAverage

Management Score, M

(1)=(2)+(3)

Reallocation effect

(Olley-Pakes, OP)

(2)

Unweighted Average

Management Score

(3)

“Deficit” in Share-

weighted Manageme

nt Score relative to

US(1)-0.67

“Deficit” in Reallocation relative

to US

(2)-0.36

% of deficit in

management score due to worse

reallocation(6)=(5)/(4)

US 0.67 0.36 0.31 0 0Sweden 0.42 0.22 0.20 -0.25 -0.14 56%Japan 0.32 0.18 0.14 -0.35 -0.18 51%Germany 0.31 0.28 0.03 -0.36 -0.08 22%Canada 0.17 0.25 -0.07 -0.50 -0.11 22%Great Britain -0.07 0.17 -0.24 -0.74 -0.19 26%Poland -0.14 0.18 -0.32 -0.81 -0.18 22%Italy -0.15 0.07 -0.23 -0.82 -0.29 35%France -0.31 0.10 -0.41 -0.98 -0.26 27%Brazil -0.34 0.24 -0.59 -1.01 -0.12 12%China -0.50 0.10 -0.61 -1.17 -0.26 22%Argentina -0.51 0.14 -0.66 -1.18 -0.22 19%Portugal -0.53 0.09 -0.62 -1.20 -0.27 22%Greece -0.98 -0.13 -0.85 -1.65 -0.49 30%

TAB 3: DECOMPOSING AGGREGATE MANAGEMENT GAP INTO REALLOCATION & UNWEIGHTED MEAN DIFFERENCE

INFORMATION: ARE FIRMS AWARE OF THEIR MANAGEMENT PRACTICES BEING GOOD/BAD?

We asked:

“Excluding yourself, how well managed would you say your firm is on a scale of 1 to 10, where 1 is worst practice, 5 is average and 10 is best practice”

We also asked them to give themselves scores on operations and people management separately

-6-4

-20

2la

bp

0 2 4 6 8 10Their self-score: 1 (worst practice), 5 (average) to 10 (best practice)

bandwidth = .8

Lowess smoother

SELF-SCORES UNCORRELATED WITH PRODUCTIVITY

Labo

r P

rodu

ctiv

ity

Self scored management

* Insignificant 0.03 correlation with labor productivity, cf. management score has a 0.295

Dep. Var. Management Self-score Management

FEs SIC3 SIC3 Firm SIC3 SIC3 Firm

Sample All 2+ obs 2+ obs All 2+ obs 2+ obs

Comp- 0.064*** 0.082*** 0.119** -0.038* -0.041 -0.046etition (0.018) (0.031) (0.051) (0.023) (0.039) (0.073)%college 0.115*** 0.109*** 0.040*** 0.069***

(0.008) (0.014) (0.011) (0.020)Ln(emp) 0.175*** 0.157*** 0.069*** 0.060***

(0.009) (0.017) (0.012) (0.022)

Obs 8,776 3,276 3,349 7,960 2,934 3,007

TABLE 11: COMPETITION AFFECTS FIRM’S SELF-PERCEPTIONS OF MANAGEMENT QUALITY

Notes: Controls include country & year dummies, public & interview noise (interviewer, time, date & manager characteristic). SEs clustered by firm.

ALGORITHM

• Discretise state space for M,K and A.• Value functions for entrants, incumbents. Numerical

contraction mapping to obtain fixed point• Investment policy correspondences (for M & K). L

defined optimally using first-order conditions• Simulate data and focus on steady states after 50 years

for 10,000 firms• Change parameters & examine different steady states

– Increase in competition– Increase in distortions

• Compare actual and simulated data

PARAMETERS

Parameter Value Rationale

ρ TFP AR(1) 0.95 Cooper and Haltiwanger (2006)

α Capital 1/3 Capital share

β Labor 1/2 Labor share

γ Management (MAT) 1/6 Assumed management share (intangibles)

b Management (MAD) 2 Assumed costs of non-optimal management

E Sunk cost 100 Assumed 100 times unit price

F Fixed cost 25 Assumed 25 times unit price

e Price elasticity 4 Markup of 25%, midpoint of literature

QK Capital quad. Acs 100 ACs about 5% of revenue, Bloom (2009)

QM Manag. quad. Acs 100 ACs about 5% of revenue, Bloom (2009)

Dτ Distortions upper bound

10% Hsieh and Klenow (2009)

TRANSITION MATRIX: MANAGEMENT, 2006-2009, 1600 firmsQuintile in 2009 Bottom Second Third Fourth TopQuintile in 2006Bottom 52 22 15 9 3

Second 23 25 25 8 10

Third 16 24 26 19 15

Fourth 7 16 26 26 24

Top 6 8 13 28 46

Quintile in 1977 Bottom Second Third Fourth Top

Quintile in 1972

Bottom 36 18 11 18 16

Second 20 19 22 22 17

Third 16 24 22 24 13

Fourth 9 7 17 35 34

Top 5 6 7 16 65

TRANSITION MATRIX: TFP 1972-1977, BAILEY ET AL (1992)

CONCERNS WITH THE CROSS-COUNTRY DECOMPOSITION

•Re-weighting for sample responses. Use alternative covariate sets (e.g. just emp in selection in Table B1)•Using just labor as size measure (because cleanest). Use capital as well to get “input index” (Table B2)•Dropped multinationals because “size” unclear: what if we include (Table B3)•Only looking at firms between 50 to 5000 what about very small and very large firms? (Table B4)

56

Figure B1: Relative Management (weighted by employment shares; only emp in selection equation)

Notes: These are the differences relative to the US of (i) the weighted average management scores (sd=1, blue bar) and (ii) reallocation effect (OP, light red bar). Domestic firms, . 2006 wave. Response bias corrections use country-specific employment only

-1.54

-.39

-1.14

-.2

-1.11

-.19

-.92

-.22

-.78

-.14

-.69

-.18

-.68

-.14

-.33

-.05

-.26

-.12-.2

-.09

00

-1.5

-1-.

50

grptcnfrpoitgbgejpswus

mean of rel_man mean of rel_OP

Figure B2: Relative Management (weighted by labor and capital inputs)


-1.52

-.36

-1.16

-.24

-1.11

-.18

-.88

-.16

-.73

-.2

-.72

-.1

-.71

-.16

-.33

-.05

-.24

-.14-.17

-.06

00-1

.5-1

-.5

0

grcnptfritpogbgeswjpus


Figure B3: Relative Management (weighted by employment shares; multinationals included)


-1.29

-.21

-1.27

-.26

-.89

-.08

-.82

-.16

-.68

-.15

-.53

-.05

-.51

-.06

-.32

-.12

-.27

-.08-.13

-.02

00

-1.5

-1-.

50

cngrptpogbitfrswgejpus


-2-1

.5-1

-.5

0

grptitpofrgbjpswgeus

mean of rel_man mean of rel_WM

Figure B4: Relative Management scores. Correcting for missing very small and very large firms.

Notes: These are the differences relative to the US of the weighted average management scores. Response bias corrections. Blue bar is baseline results and red bar corrects for missing firms with under 50 or over 5000 employees. The correlation between the two bars is 0.95.

Dependent Variable

Human Capital Management (z-score)

Fixed Capital Management (z-score)

Relative style of Management


Human Capital Management (z-score)

Fixed Capital Management (z-score)


HC FC HC-FC HC-FC HC FC HC-FC% Degree 1.193*** 0.610*** 0.583*** 0.561***(Firm) (0.072) (0.065) (0.085) (0.088)% Degree 0.743*** 0.403* 0.341**

(Industry) (0.229) (0.222) (0.167)

Ind Dums No No No Yes No No NoObs 7,641 7,641 7,641 7,641 8,833 8,833 8,833

MANAGEMENT AS DESIGN: STYLES DO DIFFER SYSTEMATICALLY ACROSS FIRMS AND INDUSTRIES

Note: OLS with standard errors clustered by 156 three digit industries below coefficients. “Human Capital Management” (HC) is the average of the z-scores of questions 13,17 and 18 (and this overall average z-scored). “Fixed Capital Capital Management” (FC) is the average of the z-scores of questions 1, 2 and 4 (and this overall average z-scored). “Relative style of management” is the simple difference of HC and FC. All columns include year dummies, ln(firm size), ln(firm age) and a full set of country dummies. “% degree (firm)” is the proportion of the firm’s employees with a college degree. “% degree (industry-level in US)” is the proportion of workers in a three digit SIC with a college degree from the US Current Population Survey.

Management in the US vs developing countries


0.5

1K

ern

el D

en

sity

Est

ima

tion

1 2 3 4 5Firm Average Management Score

India

Nicaragua

Ethiopia

Ghana

Zambia

Kenya

Tanzania

United States

Bottom 25% in US: 2.88

Note: Percentage of firms scoring within the range of the bottom quartile of US firms: 64% of INfirms, 80% of NIC firms, 94% of ET firms, 95% of GH firms, 74% of KE firms, 90% of TZ firms,77% of ZM firms

US

Performance measure

ln ln( ) ln( )it M it L it K it X it itY M n k x u

Capital

Management(z-score each question,

average & z-score again) Labor

Other controls

• M, Management Index is average of all 18 questions (sd=1)

• Other controls include: % employees with college degree, average hours worked, firm age, industry, country & time dummies & noise (e.g. interviewer dummies).

PERFORMANCE REGRESSIONS

Education also important in accounting for management differences

• Important in both MAD and MAT models:– MAT: reduces the cost of good management– MAD: likely to change the optimal choice of practices

• Empirically challenge is identifying causal effects, tried:– Variation in universities across locations by country– Using land-grant colleges in the US (Moretti, 2004)

motivation: what is management?

Documents