Information and coordination in Organizations Luis Garicano NBER, May 2013 Organizations and Coordination Arrow (1974) “Organizations are a way of achieving the benefits of collective action in situations in which the price system fails” March and Simon (1993) . “Organizations are systems of coordinated action among individuals and groups” Aims of organization • Division of labor – Creates coordination problems – Creates informational asymmetries • Resource allocation problem – Information intensive – Incentive conflicts But what does it have to do with development question • If markets in developing countries are less efficient, and market transactions less reliable, regulators more unpredictable more of the action will take place inside firms – Price system more likely to fail Nick Bloom´s Pictures of textile factories 6 Organizations must deal with two problems • Coordination Assume people were all well intentioned and share a common goal, must get over bounded rationality/limited information/limited knowledge • Motivation Need to motivate people for the common goal Coordination and motivation • Coordination – Utilization of knowledge – Communication, ransmission of information – Allocation of talent • Motivation – Informational asymmetries – Contractual Incompleteness Coordination matters • The division of labor is limited by – Extent of the market (Smith) – Coordination costs (Becker and Murphy, QJE 93) • Firms devices to reduce coordination costs (knowledge divided and integrated) • Broadly, if firms are bad at solving coordination , inefficiently little specialization Arrow (1974) on Information Requirements of Price System • Arrow-Debreu solution to competitive equilibrium requires agents to have an enormous amount of information • This is particularly clear in the case of uncertainty: – The possibility of using the price system to allocate uncertainty is limited by the structure of the information channels in existence. • What is important here is that the presence or absence of information channels is not prescribed exogenously to the economic system. – The question then is how much information to collect?: cost/benefits Transmitting information is not just about Incentive Compatibility ‘Standard’ neoclassical theory does not really worry about limits on communication and knowledge • As long as actions of members are incentive compatible, they can be implemented. Only private incentives limit information transmission • In fact, communication and knowledge acquisition is very costly Information costs: is this just “technology”? Arrow distinguishes three characteristics of information costs (1)Strongly decreasing returns -- fixed factor is the individual (2)Irreversible investment aspect Thus history matters Information costs (2) (3) Information costs are different in different directions – Things are ‘near’ in some space to others – Similarly, it is easier to communicate with some individuals – Also, complementarities with production activities Organization and Information Costs • Arrow argues that an organization can acquire more information than any individual • Can economize on the use of information, – (sufficient statistics, for example) • Thus an organization is useful for information handling – Codes – Hierarchies Here focus on coordination role of organizations Caveats • Aim is on understanding underlying organization mechanisms rather than impact on growth – Research must study that! • A lot of these mechanisms exist within firms, often also between firms Application I: ICT impact on coordination and motivation Likely to reduce both – Coordination costs (by making information cheaper, more readily available) – Motivation costs (by e.g. hardening information) Example 1. Paravisini-Schoar (2013) • A new technology for loan processing: scoring system • How does it impact productivity of bank? – Does it increase the amount of information employees have, improving decision making and solving more problems? (coordination) – Or does it help deal with incentive conflicts by making information “hard” (motivation) Example 2. Baker Hubbard, Trucking (QJE04/AER 03) • A new technology for trucks: on board computing • How does it impact productivity of truckers? – Better assignment of loads, better coordination of fleet, resulting from better information and higher capacity utilization – Less incentive conflicts on driving, lower moral hazard Example 3. Garicano-Kaplan JIE 2001 (B2B Car sales) Physical Flow Physical Flow Dealership Information Flow Dealer Dealership Dealer Auction Also Rental Car Firms and Manufacturers Lessor Bank Also Rental Car Firms and Manufacturers Lessor Bank Auction Vehicle Inventory Dealer Dealer Dealership Personnel Transport Information Systems Appraisal Company Dealer (flies in) Information Systems Lessee Dealership Dealership Leased Vehicle Personnel Transport Lessee Personnel Dealer Dealer Inventory Dealer Inventory Leased Vehicle Transport Information Systems Used Car Customer Used Car Customer Coordination 1. Utilizing knowledge optimally 2. Reducing human fallibility 3. Matching jobs with talent/job assignment 4. Corporate culture and knowledge 5. Communication and firm language 1. UTILIZATION OF KNOWLEDGE Efficiencies of Division of Labor • Sherwin Rosen: – Indivisibility of Human Capital Investment (Rosen 1983) “implies a fixed cost element independent of subsequent utilization of skills.” This implies that two identical individuals gain by specializing – the matching of people with activities for which they are suited (Rosen 1978). Hierarchy and Utilization of Knowledge • Hierarchy: A device to match problems with solutions (Garicano, JPE 2000) – When matching problems with solutions is very hard, hierarchies increase the utilization of knowledge of experts by using less knowledgeable subordinates to protect them from easy things others can do – The purpose of the hierarchy is to economize on cognitive costs; through a hierarchy, a manager can project his superior knowledge through many workers. Workers deal with routine questions and managers with exceptions Manager Span of control: number of workers reporting to manager Worker autonomy: low if managers take most decisions; high if workers take most decisions Workers 24 HOW DO FIRMS DETERMINE THEIR HIERARCIES? • Firms face tasks in the interval [0,1] distributed according to density function f(z) z0 0 Routine 1 Non-routine Delegated Tasks (Worker) Centralized Tasks (Manager) 25 • In order to solve problems, the firm needs to train each worker at a training cost a (information “acquisition”) • Asking the manager entails a communication cost h (“helping” cost), subject to the managers total time constraint • So the optimal organization will balance of training and helping costs, with z0 decreasing in “a” and increasing in “h” Knowlegdge based hierarchy (JPE, 2000) f(z) zw z1 z2 z3 z Management by Exception Asking for help versus learning • Worker deals with routine problems, manager with exceptions • Information technology, cheaper training, higher human capital decentralizes, “empowers” – Makes learning to solve a problem cheap relative to asking • Communication technology “centralizes” – Makes asking others cheaper relative to learning to solve it Examples of management by exception • Resident - attending physician • Software customer service /technical support – Orlikowski (1996) • Paravisini-Schoar (2013) loan processing • Casaburi and Kremer (2013) farmer queries • Supervisor - worker in factory floor • Partner - associate in professional services (GaricanoHubbard, 2010, 2011, 2013) Key role of middle level managers (Antras, Garicano, Rossi-Hansberg 2006) • By shielding top management from routine problems faced by host country workers, the presence of middle managers improves the efficiency of the transmission of knowledge Key empirical counterparts of theory Observable Parameters • Cost of acquiring information, affected by cost of training, IT • Cost of communicating, helping, affected by networks, phones etc. • Complexity / predictability of production Choices • Span • Number of layers of management • Proportion of decisions made below to decisions at higher levels • Level of training of lower and higher level employees • Discretion/decisions made by the higher ups Application 2: Distinguishing IT and Ct (Bloom, Garicano, Sadun, Van Reenen, 2012) Clear predictions for information technologies (IT) and communication technologies (CT) on firm organization IT will reduce information acquisition costs (a), leading to an: • Increase in z0 (decentralization) as workers can tackle more tasks • Increase in s (the span of control) as workers ask less questions so that managers can direct more people CT will reduce communication costs (h), leading to: • A reduction in z0 (centralization) as cheaper to ask for help • An ambiguous impact on the span of control, as more questions are asked but each takes less time to ask 31 2. FALLIBILITY AND ORGANIZATION Architecture and Errors • Starting point: The performance of an organization is determined by how centralized or decentralized it is • Individuals have a limited ability to absorb and process information • Two main types of architecture: – Decentralized or Flat (“poliarchy” in S&S): – Centralized (“Hierarchy “) • Question: how these two architectures affect quality of decision making? And how do we allocated talent given this? – False Positive – False negatives Fallibility of centralized and decentralized structures • Decentralized: independent screens – Accepted projects are those that were accepted by either one or the other (parallel) – Probability of acceptance: p(x ) +p(x ) (1 - p(x )) • Centralized/Hierarchical: successive screens: – accepted projects are those accepted by both. (serial). – Probability of acceptance: p(x ) * p (x) Comparison – A decentralized structure selects a larger proportion of projects than a hierarchy (Proposition 1) – The incidence of Type-I error is higher in a hierarchy (reject too much); of Type II in a decentralized structure (accept too much) • Trade off consistency versus innovativeness – Decentralized: “anything goes” advantage: we do not kill too many good projects; but we let go some bad stuff – Centralized/Hierarchy: we are cautious/we do not let in the bad stuf but we may be too selective. Empirical counterparts Parameters – Relative cost of type I and II errors • Importance of preserving reputation • Speed of change in market Choices – Number of approvals a new product or design requires – Level at which decisions are made 3. JOBS AND TASK ASSIGNMENT (A BIG RESEARCH FRONTIER) Fallibility, complementarities and substitutabilities • Production with (strong) complementarities: O Ring (Kremer QJE 93): If anything fails, all fails – Need homogeneous workforce – Allocate better workers to latter stages of production • Production with substitutability: as long as anyone knows the answer, we produce – Need stars – Heterogeneous talent, as long as some star Production, Organization and Talent • Grossman and Maggi, AER 00 – The country with a relatively homogeneous population exports the good produced by a technology with complementarities between tasks. The country with a more diverse workforce exports the good for which individual success is more important. • Grossman, JPE 04, – country with a more heterogeneous labor force exports the good that is produced by the most talented individuals. In this country, trade exacerbates the "polarization" of labor and often worsens the distribution of income. Key ideas/Empirical Counterparts • Allocation of talent should depend on – Decision making – Complementarities and substitutabilities between tasks – crucial tasks more talent – Distribution of risks in tasks- cost of errors – Uncertainty – Speed of change Vertical Allocation of Talent • Manne’s (1965) suggestion: – “Observed size distribution is a solution the problem: allocate productive factors over managers of different ability so as to maximize output” Hierarchies and Talent (Garicano Rossi Hansberg QJE 06) 1. production requires physical inputs and knowledge • problem solving required to produce 2. Individuals choose knowledge acquired 3. Individuals may work on their own or in teams • some individuals may communicate knowledge to others 4. Individuals have heterogeneous cognitive ability Application 3: Garicano, Lelarge, Van Reenen : French firm size and productivity 4. CULTURE COMMUNICATION AND CODES Culture as a cognitive construct • Crémer: the seed of an economic approach based on bounded rationality • Question: what is the use of corporate culture? – Under what conditions are firms depositories of knowledge? Cognitive aspects of culture (Crémer, 93) • Culture: part of the knowledge stock shared by a substantial portion of the employees of the firm, but not by the general population • ‘Shared specific human capital’ or ‘bounded rationality of firm”’ – Elements • A common code • A shared knowledge of facts • A knowledge of rules of behavior (like Kreps) Corporate culture and efficiency • Consider a decision by an individual that commits his firm – By knowing facts, rules, agent can predict the behaviors of others – facilitate coordination • Trade-off • Predictability vs. rigidity – Culture makes communication cheaper Empirical implications: how common the culture? A. Prat, EER 2002: Information received must be common when coordination matters – If coordination is important, diversified information will dampen the reaction to information about state. – On the other hand, if coordination is of little importance, diversified information makes it feasible for the agents to react- strongly to the information that they receive Firms and Culture • Stability? – Sunk investment aspect – Consistency is required for it to perform its function Firms and Culture (2) • Theory of the Firm? – Selective intervention possible? • No, the presence of corporate culture imposes fixed costs of intervention – and the solution may not be just a bit of integration – but none. 5. COMMUNICATION AND CODES Coding and information costs Since communication channels must be designed, the code can suitably be chosen for efficiency. • Coding weakens tendency to decreasing returns • Creates an irreversible commitment by organization – Specific investment – Stability of the code – ‘Identity’ of the organization? • Important (unresolved) issue: implications for theory of the firms – one code one firm? – to decrease difficulties of communication, outsource functions with different codes? Firms and Culture: Codes • How can we model a code? A cognitive aspect of firm culture? • What does it imply? Cremer, Garicano, Prat, QJE 07 • Develop a theory of codes, when agents communicate with several different other agents. • Derive some implications for the structure of firms, and collaboration between firms. • Use it to understand the impact of certain information costs on the structure of organizations – Integration vs. disintegration – Centralization and use of hierarchies 6. CONCLUSIONS Firms and capabilities • Current theory of the firm focused exclusively on incentive issues (subeconomy/Holmstrom; property rights GHM) ∙ • Alternative view (as in Arrow, Simon, March) sees firm as bringing more information/knowledge to bear on problems – In management parlance, firms as bundles of capabilities • Examples: – BIC had capabilities in plastic injection molding and mass marketing. Used it for entry in Disposable Pens (1958),Disposable Lighters (1973) and Disposable razors (1974). – Apple's design and marketing capabilties used for entry in PCs, Laptops, portable music players (IPod), Phones (iPhone), Tables (iPad) • Hypothesis: Understanding capabilites requires understanding interaction between bounded rationality and incentives – As in (Bob G) communicating the caterogies, the content of the relational contract Empirical implications and counterparts: Frontier • Capabilities – Knowledge utilization – Correcting errors-making best decisions – Job assignment – Culture – Common Language common codes Back to growth question: Coordination • High coordination costs lead to little specialization – Unpredictability high, high communication cost – Job assignment random (can aim to change it) does not correspond to job design characteristics – Leverage of talent limited, middle managers missing – Incompatible languages, rules, codes Motivation • Motivation – Multitask incentives • Output is less measurable? • Can we outsource less? – Careers • Without deep markets, careers matter more inside firms- more career concerns (each business works less well with less market) – Relational contracts • Lack of court enforcement requires more use of relational contract • Relational contracts perform worse- less future, more uncertainty, etc (But Bob G. opposite direction also) – Hold Up more prevalent, less punished? (Raj)