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) MAT v MAD – are management difference optimal? Ohio, USA Maharashtra, India 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 Data Management Models Testing the models Management and cross-country and firm TFP Survey methodology (following Bloom & Van Reenen (2007)) 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 Example monitoring question, scored based on a number of questions starting with “How is performance tracked?” 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 incentives question, scored based on questions starting with “How does the promotion system work?” 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 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 United States Japan Germany Sweden Canada Great Britain France Italy Australia Poland Mexico Singapore New Zealand Northern Ireland Portugal Republic of Ireland Greece Chile China Brazil Argentina India Colombia Kenya Zambia Nicaragua Ethiopia Ghana Tanzania N=1289 N=176 N=658 N=403 N=412 N=1208 N=632 N=313 N=454 N=364 N=515 N=306 N=150 N=136 N=307 N=160 N=269 N=581 N=755 N=1111 N=558 N=840 N=127 N=120 N=50 N=74 N=122 N=87 N=80 2 2.5 3 Average Management Scores, Manufacturing Africa Asia Australasia Europe Latin America North America 3.5 Note:survey Firms between 50 of and9/20/2013. 5000 employees, Raw datanot yet included in the paper Data includes 2013 wave as Africa data 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” 3.5 Average management scores across countries are strongly correlated with GDP Africa Australasia Asia United States Europe Japan Germany Sweden Canada Latin America 3 North America management x log of GDP PPP per capita United Kingdom ItalyFrance Australia Poland Mexico China Chile Brazil Singapore New Zealand Portugal Republic of Ireland Greece Argentina 2.5 India Colombia Kenya Zambia Ethiopia Ghana Nicaragua 2 Tanzania 7 8 9 10 11 Log of 10-yr average GDP based on PPP per capita GDP(Current int'l $ - Billions) Note: April 2013, World Economic Outlook (IMF) indicator Data includes 2013 survey wave as of 9/20/2013. Africa data not yet included in the paper Like TFP management also varies within countries 2 Japan 3 Germ any 4 Sweden 5 Canada 6 Great Britain 7 France 8 Italy 9 Aus tralia 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 0 .5 1 0 .5 1 0 .5 1 0 .5 1 1 United States 1 26 Nicaragua 27 Ethiopia 28 Ghana 3 4 5 29 Tanzania 0 .5 1 25 Zambia 2 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 Firm Average Management Score Graphs by rank_cty Data includes 2013 survey wave as of 9/20/2013. Africa data not yet included in the paper Management Data Management Models Testing the models Management and cross-country and firm TFP 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 Management as a technology: Francis Walker “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.” Francis Walker (QJE, April 1887) Walker ran the 1870 US Census, and was the founding president of the AEA and president of MIT Management as a technology: Alfred Marshall “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) Two perspectives on these management differences in economics • 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) 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 1 1 Result 1: Performance (size, TFP, profits) increase in management with MAT but not MAD .6 0 .2 .4 Sales 0 .2 .4 Sales .6 .8 MAD: Y=AKαLβ/(1+|M-M*|) .8 MAT: Y=AKαLβMγ 1 2 3 Management 4 5 1 2 3 Management 4 5 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). 4 4 Result 2: Average Management increases with competition (elasticity) with MAT but not MAD MAD: Y=AKαLβ/(1+|M-M*|) 3 2.5 2 1 1 1.5 1.5 2 2.5 3 mean of ilogmanagement 3.5 3.5 MAT: Y=AKαLβMγ 2 4 6 8 10 Comp increasing 8 2 4 10 6 Comp increasing 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. Result 3: Sales weighted management is higher in less distorted economies with MAT but not MAD MAD: Y=AKαLβ/(1+|M-M*|) 4.2 4.2 MAT: Y=AKαLβMγ 2 5 10 15 4.1 4 20 Distortions increasing (% sales lost) 3.8 3.8 3.9 India 3.9 4 mean of ilogmanagement 4.1 US 2 5 10 15 20 Distortions increasing (% of sales lost) 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. 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) Measuring Data Management Models Testing the models • Performance • Competition • Reallocation Management and cross-country and firm TFP TABLE 2: Performance is robustly correlated with management (consistent with MAT) Dependent variable Ln(sales) Ln(sales) Ln(employ Profit rate 5yr Sales -ment) ROCE growth Exit Exit OLS Fixed Effects OLS 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 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. TABLE 3: Performance: use the Great Recession as a natural experiment and find again well managed firms perform better Growth in firm sales Dependent variable Shock (Industry Sales) -0.033** (0.014) -0.035** (0.014) 0.027** (0.011) Management*SHOCK -0.051*** (0.014) Management 0.002 (0.006) -0.014 (0.010) 0.001 (0.006) -0.052*** (0.014) 0.018* (0.010) -0.008 (0.009) Firms Observations 1,567 1,653 1,567 1,653 1,599 1,685 1,599 1,685 Shock (Industry Trade) Management*SHOCK 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 Table 4: Contingent performance: not consistent with MAD (best to be at optimal industry/country point) Dependent Variable Ln(Sales) Ln(Sales) Ln(Sales) Cell SIC3 ×Country Country 0.236*** (0.032) 0.226*** (0.027) Management *I(M below cell ave) Management *I(M above cell ave) F test of symmetry above v below cell ave Cell Clusters Observations ROCE Growth Exit SIC3 SIC3 ×Cty SIC3 ×Cty SIC3 ×Cty 0.219*** (0.034) 0.219*** (0.030) 0.226*** (0.039) 0.223*** (0.030) 1.681*** (0.528) 1.679*** (0.436) 0.038 (0.028) 0.044* (0.024) -0.018** (0.005) -0.014** (0.004) 0.230 0.984 0.785 0.990 0.388 0.060 796 8,003 18 8,314 177 8,292 900 8,793 921 8,007 1,137 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. Measuring Data Management Models Testing the models • Performance • Competition • Reallocation Management and cross-country TFP TABLE 5: Competition improves management (consistent with MAT) Competition proxies Import penetration Management (estimated in levels) 0.805*** (0.236) 1- Lerner Index1 0.608*** (0.230) 17.53* (3.85) # of competitors Balanced panel Obs Management (estimated in differences) 20.68** (6.65) 0.120** (0.052) 0.121*** (0.023) No No No Yes Yes Yes 2,657 2,819 2,789 412 429 432 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 Measuring Data Management Models Testing the models • Performance • Competition • Reallocation Management and cross-country TFP Table 6: More reallocation to better managed firms in the US where markets generally less distorted (consistent with MAT) Dependent Variable Management (MNG,US base) MNG*Argentina MNG*Australia MNG*Brazil MNG*Canada MNG*Chile MNG*China MNG*France MNG*Germany MNG*Greece MNG*India MNG*Ireland MNG*Italy MNG*Mexico MNG*New Zealand MNG*Japan MNG*Poland MNG*Portugal MNG*Sweden MNG*UK Observations Employees 201.9*** 5,842 Employees 359.7*** -270.9** -259.3* -211.7* -169.3 -92.6 -84.9 -489.5** -9.0 -355.9*** -145.4 -258.8** -283.1*** -250.1* -375.7* -297.3** -308.1*** -308.9*** -228.7* -125.1 3,858 Sales Growth 0.092*** -0.134*** -0.145** -0.101*** -0.131** -0.150 -0.060 -0.085* -0.080* -0.089** -0.066 -0.085 -0.092** -0.075* 0.718*** -0.099** -0.058 -0.109** -0.068 -0.054 2,756 Notes: Includes year, country, three digit industry dummies, # management questions missing, firm age, skills and noise controls. Table 7: Reallocation also stronger in countries with lower labor and trade regulations (consistent with MAT) Dependent Variable: Management (MNG) Management*Employment Protection (World Bank Country Index) Employment 231.46*** 356.73*** 110.97* (37.12) (55.89) (66.30) -1.43** (0.69) Management*Trade Costs (World Bank Country Index) (0.69) -0.18*** (0.05) Tariffs (country x country) Management*Tariff Observations 5,760 5,017 -4.96 (4.12) -8.25** (3.35) 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). Measuring Data Management Models Testing the model • Performance • Competition • Reallocation • Extensions: skills Management and cross-country TFP Management and Education: use UNESCO World Higher Education Database university locations (N=9,081) Distance to the nearest university seems to matter for firm skills and management Dependent Variable: Drive time to nearest university Manage Manage ment % firm employees with degree ment ment OLS OLS OLS IV -0.049*** -1.534*** (0.019) (0.423) 0.789*** 3.190*** (0.082) (1.113) 6,406 6,406 Manage % employees with degree in the firm Observations 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. Measuring Data Management Models Testing the model • Performance • Competition • Reallocation • Extensions: skills 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 First calculate the employment weighted difference in management (from the US as baseline) -.14 -.25 -.08 -.18 -.5 -.35 -.11 -.19 -.12 -.18 -.29 -.36 -.26 -.26 -.22 -.27 -.49 -.49 -.81 -.83 -1 -.74 -.98 -1.02 -1.18 -1.19 -1.2 -1.5 Management gap with the US 0 0 -2 -1.65 us sw jp ge ca gb po Reallocation mean of rel_OP it fr br cn ar pt gr Within of rel_zman mean Firm 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) -.14 -.25 -.08 -.18 -.5 -.35 -.11 -.19 -.12 -.18 -.29 -.36 -.49 -.81 -1 -.74 -.26 -.26 -.22 -.27 30% of US-Greece management gap due to better US -.83 reallocation -.98 -.49 -1.02 -1.18 -1.19 -1.2 -1.5 Management gap with the US 0 0 -2 -1.65 us sw jp ge ca gb po Reallocation mean of rel_OP it fr br cn ar pt gr Within of rel_zman mean Firm 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 Second, estimate impact of management on TFP using result from micro regressions and field experiments result: ↑1 SD management ≈ ↑ 10% TFP Country US Sweden Japan Canada Great Britain Italy France Brazil China Argentina Portugal Greece Unweighted av. Share-Weighted Average Management Deficit with US 0 -0.25 -0.35 -0.50 -0.74 -0.81 -0.82 -0.98 -1.01 -1.17 -1.18 -1.65 TFP GAP with US Proportion of TFP gap due to Management 32.2 33.6 22.3 20.3 17.2 25.3 59.6 78.3 57.3 24.9 51.0 7.8% 10.4% 22.4% 36.5% 47.7% 38.7% 16.9% 14.9% 20.6% 48.2% 32.4% 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). Interestingly, also get similar 25% share for cross firm contribution of management to TFP spread • 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 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: The difficulties of defining ownership in Europe 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” ?” Americans on geography Interviewer: “How many production sites do you have abroad? Manager in Indiana, US: “Well…we have one in Texas…” MY FAVOURITE QUOTES: Don’t get sick in Britian 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 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” MY FAVOURITE QUOTES: Don’t get sick in India Interviewer : “Do you offer acute care?” Switchboard: “Yes ma’am we do” 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” “OLLEY PAKES” (OP) DECOMPOSITION OF WEIGHTED AVERAGE MANAGEMENT SCORE (M) IN GIVEN COUNTRY Overall management z-score of firm i Size of firm i M M iYi i [( M i M i )(Yi Yi )] M i OP M Covariance (Olley-Pakes, 1996, reallocation term) Unweighted mean of management score DECOMPOSING THE RELATIVE MANAGERIAL DEFICIT BETWEEN COUNTRY j AND THE US ECONOMY M M k US Difference in aggregate share-weighted Management scores (OP OP ) ( M M ) k US Difference in reallocation (between firm) k US Difference in unweighted Means (within firm) TAB 3: DECOMPOSING AGGREGATE MANAGEMENT GAP INTO REALLOCATION & UNWEIGHTED MEAN DIFFERENCE Country US Sweden Japan Germany Canada Great Britain Poland Italy France Brazil China Argentina Portugal Greece ShareWeighted Average Managemen t Score, M Reallocation effect (Olley-Pakes, OP) Unweighted Average Management Score (1)=(2)+(3) 0.67 0.42 0.32 0.31 0.17 -0.07 -0.14 -0.15 -0.31 -0.34 -0.50 -0.51 -0.53 -0.98 (2) 0.36 0.22 0.18 0.28 0.25 0.17 0.18 0.07 0.10 0.24 0.10 0.14 0.09 -0.13 (3) 0.31 0.20 0.14 0.03 -0.07 -0.24 -0.32 -0.23 -0.41 -0.59 -0.61 -0.66 -0.62 -0.85 % of deficit “Deficit” in “Deficit” in in ShareReallocatio manageme weighted n relative nt score Manageme to US due to nt Score worse relative to reallocation US (2)-0.36 (6)=(5)/(4) (1)-0.67 0 0 -0.25 -0.14 56% -0.35 -0.18 51% -0.36 -0.08 22% -0.50 -0.11 22% -0.74 -0.19 26% -0.81 -0.18 22% -0.82 -0.29 35% -0.98 -0.26 27% -1.01 -0.12 12% -1.17 -0.26 22% -1.18 -0.22 19% -1.20 -0.27 22% -1.65 -0.49 30% 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 labp Labor Productivity -4 -2 0 2 SELF-SCORES UNCORRELATED WITH PRODUCTIVITY Lowess smoother 0 2 4 6 8 Their self-score: (worst management practice), 5 (average) to 10 (best practice) Self 1scored 10 bandwidth = .8 * Insignificant 0.03 correlation with labor productivity, cf. management score has a 0.295 TABLE 11: COMPETITION AFFECTS FIRM’S SELFPERCEPTIONS OF MANAGEMENT QUALITY Dep. Var. Management Self-score Management FEs SIC3 SIC3 Firm SIC3 SIC3 Firm Sample Competition %college All 0.064*** (0.018) 0.115*** (0.008) 0.175*** (0.009) 2+ obs 0.082*** (0.031) 0.109*** (0.014) 0.157*** (0.017) 2+ obs 0.119** (0.051) All -0.038* (0.023) 0.040*** (0.011) 0.069*** (0.012) 2+ obs -0.041 (0.039) 0.069*** (0.020) 0.060*** (0.022) 2+ obs -0.046 (0.073) 8,776 3,276 3,349 7,960 2,934 3,007 Ln(emp) Obs 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 10% Hsieh and Klenow (2009) Distortions upper bound TRANSITION MATRIX: MANAGEMENT, 2006-2009, 1600 firms Quintile in 2009 Quintile in 2006 Bottom Bottom Second Third Fourth Top 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 TRANSITION MATRIX: TFP 1972-1977, BAILEY ET AL (1992) Quintile in 1977 Bottom Second Third Fourth Top 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 Quintile in 1972 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) 0 0 0 -.05 -.09 -.12 -.14 -.2 -.18 -.14 -.22 -.26 -.19 -.2 -.33 -.5 -.39 -.68 -.69 -.78 -1 -.92 -1.14 -1.5 -1.11 -1.54 us sw jp ge gb mean of rel_man it po fr cn pt gr mean of rel_OP 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 Figure B2: Relative Management (weighted by labor and capital inputs) 0 0 0 -.05 -.06 -.17 -.14 -.1 -.16 -.16 -.2 -.24 -.18 -.24 -.33 -.5 -.36 -.71 -.72 -.73 -1 -.88 -1.16 -1.5 -1.11 -1.52 us jp sw ge gb mean of rel_man po it fr pt cn gr mean of rel_OP 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 Figure B3: Relative Management (weighted by employment shares; multinationals included) 0 0 0 -.02 -.13 -.06 -.08 -.05 -.12 -.16 -.26 -.21 -.32 -.5 -.27 -.08 -.15 -.51 -.53 -.68 -.82 -1.27 -1.29 gr cn -1.5 -1 -.89 us jp ge sw fr mean of rel_man it gb po pt mean of rel_OP 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 -2 -1.5 -1 -.5 0 Figure B4: Relative Management scores. Correcting for missing very small and very large firms. us ge sw jp gb mean of rel_man fr po it pt gr mean of rel_WM 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. MANAGEMENT AS DESIGN: STYLES DO DIFFER SYSTEMATICALLY ACROSS FIRMS AND INDUSTRIES Dependent Variable % Degree (Firm) % Degree (Industry) Ind Dums Obs Human Capital Managem ent (zscore) HC 1.193*** (0.072) No 7,641 Fixed Capital Manage ment (zscore) FC 0.610*** (0.065) No 7,641 Relative style of Managem ent Relative style of Managem ent HC-FC 0.583*** (0.085) HC-FC 0.561*** (0.088) No 7,641 Yes 7,641 Human Capital Managem ent (zscore) HC Fixed Capital Manage ment (zscore) FC Relative style of Manage ment 0.743*** (0.229) 0.403* (0.222) 0.341** (0.167) No 8,833 No 8,833 No 8,833 HC-FC 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 1 India Nicaragua Ethiopia Ghana Zambia Kenya Tanzania .5 United States Bottom 25% in US: 2.88 0 US 1 2 3 4 Firm Average Management Score 5 Note: Percentage of firms scoring within the range of the bottom quartile of US firms: 64% of IN firms, 80% of NIC firms, 94% of ET firms, 95% of GH firms, 74% of KE firms, 90% of TZ firms, 77% of ZM firms Data includes 2013 survey wave as of 9/20/2013. Africa data not yet included in the paper PERFORMANCE REGRESSIONS Performance measure ln Yit M M it L ln(nit ) K ln(kit ) X xit uit Management (z-score each question, average & z-score again) Other controls Labor Capital • 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). 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)
© Copyright 2024 Paperzz