Boundaries of the firm
A Socratic walk-through of the boundaries of the firm — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why does a company hire employees instead of buying every task on the open market?
Economics spends most of its effort explaining why markets allocate well: prices carry information, competition disciplines, and nobody has to be told what to do. Take that seriously and a puzzle appears immediately. If the market is such a good coordinator, why is so much economic activity conducted inside organisations where prices are absent and people are simply instructed?
Ronald Coase asked this in 1937, and it is genuinely awkward: within a firm the price mechanism is switched off on purpose. A designer does not bid for the accountant's time. So either firms are an inefficiency we tolerate, or the market has costs we have not been counting.
Reasoning it through
REASONING #Try running a firm without one. You need a component made. Instead of employing anyone, you go to the market each time.
What does that cost? Not just the price. You must discover who can make it and what they charge; specify it precisely enough to be contractible, which for anything subtle is hard; negotiate; monitor delivery; and when it is wrong, have some way of establishing that, possibly a court, expensively and slowly. Do it once and it is a nuisance. Do it forty times a day, with terms shifting each time, and it is the dominant cost of the business.
Now do the same thing inside. You employ someone and, within limits, you tell them what to do. The whole apparatus of specification, negotiation and enforcement is replaced by an instruction. That is Coase's insight: a firm exists because directing beats contracting for some transactions. Employment is not a purchase of a defined output but of the right to decide later — exactly what you want when you cannot say in advance what you will need.
Push it the other way, because the argument has a hole. If instruction is cheaper, why is the whole economy not one firm? Because internal coordination has its own rising costs: more layers, worse information reaching the top, weaker incentives when reward is detached from performance, and the politics that appear whenever resources are allocated by argument rather than price. Coase's answer is therefore a margin — the firm expands until the cost of organising one more transaction internally equals the cost of buying it, and the boundary is where the two curves cross.
Satisfying, but under-specified. Which transactions land on which side? Oliver Williamson's answer turns on a property of the asset. Suppose you need a die tooled specifically for your product. Once made it is worth a great deal to you and almost nothing to anyone else — and the moment the supplier has sunk that cost, you are locked to each other. Either party can now hold the other up, renegotiating at the worst moment knowing the alternative is worthless. Both foresee this, so both under-invest in the very specificity that would have been valuable. Where assets are specific, transactions frequent, and the future too uncertain for a contract to enumerate, integration is the way out; where the good is generic and traded by many, the market is fine.
There is a further turn. Grossman, Hart and Moore asked what ownership is, and answered: the residual control rights — authority over anything the contract did not cover. Whoever owns the asset holds those rights, and so has stronger incentive to invest in it and weaker incentive to invest in what the other party owns. Integrating therefore does not abolish the hold-up problem; it reassigns it, which is why integration is sometimes exactly the wrong move.
The analogy
THE ANALOGY #Think of a household deciding what to cook and what to order in. Ordering is a market transaction: you pay a price, you specify nothing, you carry no equipment. Cooking is the firm: you own the kitchen, you decide at seven o'clock what happens, and no order needs writing. Households cook the meals that are frequent, hard to specify to a stranger, and dependent on kit already owned — and they order the ones that are occasional, standard, and better made by someone with a proper oven.
a household's kitchen does not argue back, whereas a firm's internal costs are mostly human — misaligned incentives, information lost travelling up a hierarchy, and politics — so a firm's limit to growth arrives long before it runs out of counter space, and for reasons that have no analogue in the kitchen.
Clarifying the model
THE MODEL #The theory is usually stated as a choice between two clean options. In practice the interesting territory is the middle: long-term contracts, joint ventures, franchising, preferred-supplier relationships with shared engineers. These exist because they buy some of the safety of integration without its costs, and any honest account treats make-or-buy as a spectrum of governance forms rather than a switch.
The evidence needs care too. The most-repeated illustration — General Motors acquiring Fisher Body after a dispute over specific dies — has been challenged in detail by later historical work, including by Coase himself, and is no longer safe to cite as a clean demonstration. The broader pattern, that vertical integration rises with asset specificity, is much better supported than any single anecdote.
Finally, note what the framework predicts about our own moment. If contracting costs fall — through better communication, standardised interfaces, reputation systems, modular components — the crossing point moves and firms should shrink and outsource more. That is partly what we observe: contract manufacturing, outsourced payroll and logistics, work fragmenting across supplier chains. Only partly, though, since the same period produced enormous platform firms, where returns to scale, data and network effects push the other way. Transaction costs set the boundary; they are not the only force acting on it.
A picture of it
THE PICTURE #How to readMove right as the asset becomes worth less to anyone but this partner, and up as the output becomes harder to write down and verify. Bottom-left is where markets work best — cleaning and hosting are generic, checkable and endlessly re-tenderable, so owning them buys nothing. Top-right is where both problems bite at once and integration is the usual answer: custom tooling and core research each combine a partner-specific investment with an output no contract can fully specify. The off-diagonal quadrants explain the hybrids — a bottling contract is specific but measurable, so a long agreement suffices, while brand strategy is generic in assets yet hard to judge, so the contract itself must carry the weight.
What became clearer
WHAT CLEARED #A firm is not the opposite of a market but a rival way of coordinating, chosen transaction by transaction because instruction is sometimes cheaper than contracting. The boundary sits where those two costs cross, and it moves with the specificity of the assets, the frequency of exchange, and how much of the future can be written down in advance. Integrating does not eliminate the underlying bargaining problem — it reassigns the residual right to decide, which is why the question is never "should we control this?" but "whose investment would we rather protect?"
Where to go next
ONWARD #- Why some tasks are governed by franchising or joint venture rather than either pure form.
- How the same logic explains which activities a platform keeps and which it lets others build.
Key terms
TERMS #| Term | What it means |
|---|---|
| Transaction cost | the cost of arranging an exchange rather than of the thing exchanged: search, specification, negotiation, monitoring and enforcement. |
| Asset specificity | the extent to which an investment loses value outside a particular relationship, which is what creates lock-in. |
| Hold-up problem | the risk that a partner renegotiates opportunistically once you have sunk a specific investment, and the under-investment that anticipating it causes. |
| Residual control rights | the authority to decide matters the contract left unspecified; in the property-rights account, this is what ownership means. |
Every term the collection defines is gathered in the glossary.