Measuring & explaining organizational practices

  • View

  • Download

Embed Size (px)


Measuring & explaining organizational practices Nick Bloom (Stanford & NBER) based on work with Raffaella Sadun (HBS) & John Van Reenen (LSE) MIT/Harvard Org Econ Lecture 2 (February 2010). - PowerPoint PPT Presentation

Text of Measuring & explaining organizational practices

  • Main measure averages the z-score (scores normalized to mean 0, standard-deviation 1) of each variable:Hiring senior employees (discrete, 1 to 5)Maximum Capital expenditure (continuous, in $)Introduction of new products (discrete, 1 to 5)Sales and marketing (discrete, 1 to 5)

    The Bloom, Sadun and Van Reenen (2009) empirical decentralization measure

  • Decentralization varies across countriesMost centralized AsiaSouthern EuropeLeast centralizedScandinavian countries Anglo-Saxon countriesDecentralization measure

  • Decentralization also varies across firmsDecentralization measure (higher number is more decentralized)

  • External validation country level (1/2)Do these cross-country values look sensible?

    Only prior firm decentralization measure to cross-check against we are aware of is from Hofstede (1980)Surveyed c.100,000 IBM employees across 50 countries during the 1970s & early 1980sQuestions on management style (autocractic/paternalistic or consultative) and preferences for delegationCombined into Power Distance index (1-100), low means limited (preference for) delegation

  • Power distance seems correlated with our firm decentralizationDecentralizationPower distanceCorrelation= 0.80

  • External validation country level (2/2)There is also a cross-country index of Fiscal Decentralization from Arzaghi and Henderson (2005, JPubE)

    Index of Fiscal Decentralization based on 9 factors including:Government structure (e.g. unitary v federal)Local (regional/municipal) democratization & autonomyLocal (regional/municipal) control over taxation and spending (education, police, transport etc.)

    Surveyed every country with >10 million people (in 1995)

  • Fiscal decentralization is also correlated with firm decentralizationFirm DecentralizationFiscal DecentralizationCorrelation= 0.83

  • Correlation between 1st and 2nd interviews (72 firms)Internal validation the re-rater surveycorrelation 0.51 (p-value
  • General modeling frameworkPrincipal-agent Principal is the Corporate Head Quarters (CHQ)Agent is the plant manager

    Optimal decentralization depends on trade-off between:Managers typically have better local information than CHQManagers incentives diverge from firms (agency problem)

    This can of course be extended in many ways for example:Need for coordination (Alonso et al. 2008)Incentives to communicateMulti-level agency problems with CEO and owners

  • Trust and decentralizationTrust may affect optimal decentralizationFacilitate cooperative solutions in repeated game settings: e.g. Baker, Gibbons and Murphy (1999) Proxy the congruence of incentives: e.g. Aghion and Tirole (1997)More broadly, reliability of manager and/or information: Rajan & Zingales (2002), Hart & Moore (2005)

    We find evidence of robust positive relationship between trust in region where plant is located and decentralization

  • Measure trust using the World Value Survey, from the question: Generally speaking, would you say that most people can be trusted or that you cant be too careful in dealing with people? Trust by region of the country defined as % of people answering yes to first part of the trust question

    Experimental studies show this question linked with trust/trusting behavior (Glaeser et al, 2000, Sapienza et al, 2007)

    Used in prior social capital literature: e.g. Knack & Keefer (1997); Guiso, Sapienza, Zingales (2004);Measuring trust

  • Trust and decentralizationNotes: Other controls are SIC3 dummies, noise Controls (interviewer dummies, Interviewee tenure and seniority, etc.), public Listing, CEO onsite, plant size, regional GDP/head, Regional population, domestic multinational. Weighted by % of WVS respondents in region in country.SE clustered by 112 regions.

    Trust (region)1.196***0.825***0.732**(0.429)(0.290)(0.298)Rule of law (country)0.473***(0.102)No. Competitors Plant Skills0.094***(0.016)Firm Size0.044*(0.021)Plant Size0.091***(0.029)Observations354935493549Country dummiesnonoyesOther controlsnonoyes

  • Use multinationals as a second test for trustCould worry about bias due to trust proxying for other country/regional variables

    So look at affiliates of foreign multinationals and investigate whether trust in their home country also matters

  • Example A: Domestic Firm2 Sites, Single PlantD, Decentralization

  • Plant 1(Lund Site)Global HQ(Tokyo Site)French CHQ(Paris Site)Example DJapanese MNESweden CHQ(Stockholm Site)Plant 2 (Lyon Site)Do not observe DObserve DObserve D

  • Decentralization and trust: multinationals

    Sample:Multinational FirmsTrust (region of location)0.6090.5630.446(0.592)(0.843)(1.908)Trust (country of origin)0.749***0.698***0.152(0.301)(0.331)(0.152)Trust (bilateral from origin cty to location cty)1.809***2.101***(0.768)(1.035)Full set of controlsYesYesYesYesYesRegional dummiesNoNoYesNoYesCountry origin dummiesNoNoNoNoYesClusteringRegionRegionOrigin countryOrigin countryOrigin countryObservations867867867280280

  • Competition and decentralization basic theoryPrior work had found a strong positive link Guadalupe and Wulf (2008) look at Canadian free trade experiment

    But theory is actually ambiguousCompetition may affect information:Improves the value of timely responses to local conditions (e.g. Aghion & Tirole, 1997) But, reduces value of local information as more firms for the principal to learn from (e.g. Acemoglu et al. 2007) Competition may also affect incentives:Lower risk of manager abusing autonomy as incentives more aligned with firm (e.g. Schmidt 1997, Vives 2005)Less incentive to co-ordinate prices (Alonso et al. 2008)

  • Decentralization higher with more competitionNotes: Other controls are SIC3 dummies, 12 country dummies, noise controls (interviewer dummiesInterviewee tenure and seniority, etc.), public listing, CEO onsite, plant size, Number of competitors (0=none, 1=between 1 and 4, 2=5 or more (as reported by plant manager).

    Import Penetration0.131***0.184***(0.050)(0.073)1 Lerner Index6.537***2.265***(1.176)(1.081)Number of competitors0.134***0.094**(0.036)(0.034)Plant Skills0.081***0.090***0.090***(0.018)(0.016)(0.016)Ln(Firm Size)0.076**0.068***0.066***(0.026)(0.017)(0.018)Ln(Plant size)0.119**0.091**0.090**(0.024)(0.022)(0.022)Observations2,4972,4973,5873,5873,5873,587Country & Ind. dummiesnoyesnoyesnoyesClusteringCty *Sic2Cty *Sic2Cty *Sic3Cty *Sic3FirmFirm

  • Industry volatility and ageAcemoglu, Aghion, Le Large Van Reenen and Zillabotti (2007, QJE) develop a learning model

    Firm adopts/develops a new technology agent (manager) is informed about useful of technologyprincipal (CEO) is correctly aligned with owners incentives

    Predictions are that in more volatile and younger industries technology more uncertainty so more decentralization

    AALVZ indeed find strong evidence that industry volatility (SD of levels and changes of TFP) and firm age decentralization

  • Other factorsCollection of other factors which have robust signs in AALZV (2007) and BSV (2009)

    Size: larger firms and larger plants robustly more decentralization

    Skills: firms with more education employees and more education managers more decentralized

    Multinationals: controlling for size and industry, still more decentralized

  • Measuring organization: spans and decentralization

    Differences in organization across firms, countries & time

    Factors driving the organization of firms

    Organization and productivity

    ConclusionsLecture 2: Overview

  • Two channels for the impact of decentralization (1/2)Firm size:Early work on the structure of firms argued that decentralization was critical for large firms, Penrose (1959) & Chandler (1962)

    Indeed, see that larger firms are more decentralized

    Essential for productivity growth as reallocation - which accounts for of US TFP growth - needs productive firms to grow

    Also important in development as low productivity due to lack of reallocation as too few large firms: e.g. Banerjee & Duflo, 2004; Hsieh & Klenow (2008) Pawasutipaisit & Townsend (2008)

    Hence, factors driving decentralization trust, rule of law, competition also drive growth via facilitating decentraliztion

  • Two channels for the impact of decentralization (2/2)Firm level productivity:Typically think of decentralization as a control variable firms choose the right level so not right or wrong level

    But this level may be hard to change, so can be right or wrong level in the short-run if situation changes

    Good example is IT changed information flow around the firm, so changed optimal decentralization, so not all firms optimal

    Bresnahan, Brynjolfsson and Hitt (2002, QJE); Bloom, Sadun and Van Reenen (2009, NBER) find strong complementarity between IT and decentralization

    *Summary Decentralization probably key organizational trait of firms

    Varies by country Northern Europe and North America decentralized, Southern Europe and Asian centralized

    Varies by firm high trust, strong rule-of-law, tough competition, rapid industry change and education all linked to decentralization

    Essential for macro growth: firms need to decentralize to grow, and firm growth required for productivity enhancing reallocation

    Important for firm level productivity of certain factors e.g. IT

    *My five outstanding research questions

    What fraction of the differences in TFP across firms and countries can organizational differences explain?

    What are the key factors causing difference in organizations?

    How quickly can firms change real organizational structures?

    Is the optimal organizational structure chan