Summary of "Limits of Organization (Fels Lectures on Public Policy Analysis)"

Core Idea

Organizations exist because the price system fails to handle uncertainty, externalities, coordination problems, and distributive conflict.
Arrow’s central claim is not that organizations eliminate these failures, but that they replace them with new limits: authority, information costs, and commitment problems.
The book’s deepest theme is the persistent tension between individual rationality and social rationality, with no perfect institutional solution.

Price System, Efficiency, and Its Limits

Arrow treats rational action as a balance between values and opportunities, with scarcity forcing choice.
The price system is valuable because it can produce Pareto efficiency while requiring relatively little knowledge from individuals.
That apparent freedom is incomplete, since income and choice are shaped by inherited resources, abilities, and social arrangements.
The market also fails ethically and technically when it rewards selfish motives and cannot price important goods well.
Central examples of market failure include pollution, road congestion, and especially nonmarket goods like trust, loyalty, and truth-telling.
Arrow stresses that the market does not itself justify a distribution of income, so distributive justice requires standards beyond efficiency.

Uncertainty, Information, and Why Organizations Form

A major reason for organizations is the failure of contingent markets under uncertainty; in theory one could contract for every state of the world, but in practice that is too complex.
Real-world contracting is further undermined by moral hazard and adverse selection, where hidden action or hidden risk differences distort outcomes.
Following Roy Radner, Arrow treats information structure as decisive: who knows what, when, and how that knowledge can be communicated shapes what can be delegated or insured.
Information is understood qualitatively as a change in beliefs after a signal, not as a Shannon-style quantity.
Information has three salient cost properties: it is scarce as a human input, it has a large irreversible capital component, and its cost depends heavily on direction and prior familiarity.
Learning a code, scientific vocabulary, or foreign language is an information investment: costly to acquire, hard to undo, and cheap to use once mastered.
These asymmetries make organizations path-dependent, because once a communication channel or code is established, later choices are shaped by earlier investments.
Organizations manage an agenda by dividing issues into active, monitored, and passive items, depending on information costs and decision value.
An issue moves onto the agenda when signals become cheaper, when a crisis raises the value of action, or when information in one area unexpectedly reveals something relevant elsewhere.
Organizations can gather more information than individuals by splitting experiments across members, but then face the problem of communication and selective retransmission.
Internal communication works best when raw data can be compressed into sufficient statistics without losing decision value.
Coding enables coordination but also creates identity and rigidity, since different organizations develop different internal languages and habits.
This yields a trade-off: larger organizations can process more specialized knowledge, but scale and specialization eventually raise communication costs and reduce adaptability.

Authority, Responsibility, and Control

Organizations often rely on authoritative allocation, where some decide and others carry out decisions.
Arrow distinguishes personal authority from impersonal authority such as legal rules and codes.
The functional case for authority is that it economizes on information transmission when joint action is more productive than separate action.
His Hobbesian logic runs through four steps: activities interact, information is dispersed, communication is costly, and central decision-making may therefore be cheaper than full consultation.
Traffic control illustrates why authority, or even better impersonal rules, can outperform decentralized bargaining.
The main alternative to authority is consensus, but consensus works mainly when interests and information are nearly identical.
Authority cannot rest on sanctions alone, because obedience collapses if enough people defect, as in strikes, prohibition, or other failed controls.
Authority is more plausibly sustained by convergent expectations: people obey because they expect others to obey, so authority functions as a coordinating signal.
This makes authority fragile and explains why symbols, visibility, and ritual help stabilize it.
Arrow’s defense of responsibility is functional as well as moral: authority without responsibility ignores information scattered throughout the organization and makes avoidable mistakes.
Informational overload is the central reason authority needs checks, because no leader or small group can process everything relevant.
He identifies four mechanisms of responsibility: responsibility to a higher authority, to an occasional authority like voters or stockholders, to a special authority like courts, and to nonauthoritative groups such as review commissions or ombudsmen.
The preferred mechanism is intermittent review—periodic, sampled, or exception-based oversight—because continuous review can simply become another authority layer.
Modern organizations therefore need not only hierarchy, but a circulation of elites and a circulation of information and decision rules so new knowledge can enter and old codes can be revised.

What To Take Away

Efficiency is not justice, and the price system is only one institutional tool among others.
Organizations arise because markets cannot fully handle uncertainty, externalities, and dispersed knowledge, but organizations generate their own information and authority problems.
The core institutional problem is balancing authority with responsibility without destroying the coordination benefits of hierarchy.
Arrow’s most durable insight is that social systems are shaped as much by information costs and communication structures as by incentives or formal rules.