The Nature of the Firm

On 10 October 1932 a 21-year-old lecturer in Dundee wrote to Ronald Fowler about the first lecture of a course he had just begun. “As it was a new approach (I think) to this subject, I was quite pleased with myself,” he reported. “One thing I can say is that I made it all up myself.” He had asked his students why firms exist at all in an economy where prices are supposed to do the organising, and had given them an answer: “I found the reason in the costs of conducting these marketing transactions.” Then he asked Fowler: “What do you think of the idea?”¹

The writer was Ronald Coase. Five years later the argument of that lecture appeared in the journal Economica as “The Nature of the Firm”: 20 pages, 51 footnotes, one diagram and not one equation (own count)². In 1991 it was one of the two articles for which he received the Nobel prize in economics³. Crossref, the registry of scholarly references, counts more than 15,000 citations of it⁴. Yet in 1970 its author called it “much cited and little used”⁵, and in 2009, at 98, he said he was surprised that it was cited so much, since it was “little more than an undergraduate essay”⁶.

Three numbers

21

his age when he first gave the argument, in a lecture at Dundee

October 1932; the paper followed five years later

0

times the words “transaction cost” appear in the paper

it speaks of “marketing costs”, eight times

54 years

between the paper and the Nobel prize

1937 and 1991

The puzzle he took to America

Coase was born in 1910 in Willesden, a suburb of London, the only child of a Post Office telegraphist. Both his parents had left school at twelve³. He went to the London School of Economics in 1929 to read for a commerce degree designed, as he put it, for future works managers, “a choice of occupation for which I was singularly ill-suited”. Five months before his final examinations he began to attend the seminar of Arnold Plant, the new professor of commerce⁷.

Plant taught him what economists mean by coordination through prices. Plant quoted Sir Arthur Salter: “The normal economic system works itself.” Nobody is in charge of it and nobody needs to be, because prices rise where goods are scarce and producers follow them. It was the Depression, schemes for planning industry were in fashion, and Plant opposed them all. Competition working through prices, he said, would do all the coordinating that was needed⁷.

Coase accepted this and saw a gap in it. Economists also spoke of a factor of production called management, whose job was to coordinate. “Why was it needed if the pricing system provided all the co-ordination necessary?” The same question had a larger form. The Russian Revolution was fourteen years old. Lenin, as Coase remembered it, had said that the Russian economy would be run as one big factory, and economists in the West were arguing that this was impossible. Yet the West was full of factories, some of them very large⁷.

He took the question to the United States in 1931, on a travelling scholarship, to study why industries were organised so differently from one another. He went to few classes. He visited General Motors and Ford in Detroit, the steel works at Gary, Sears Roebuck and the Chicago stockyards. He spent a day in the office of a purchasing agent listening to his telephone calls, and he read the yellow pages, “where I was fascinated to find so many specialist firms operating within what we thought of as a single industry”. The literature on his subject he judged, in a letter home, to be “absolute bilge”¹.

The paper he published in 1937 puts the puzzle without Lenin or the yellow pages. Inside a firm, it observes, prices do not move people about: “If a workman moves from department Y to department X, he does not go because of a change in relative prices, but because he is ordered to do so”². It borrows an image from the Cambridge economist D. H. Robertson, for whom firms were “islands of conscious power in this ocean of unconscious co-operation like lumps of butter coagulating in a pail of buttermilk”⁸. And it asks why, if prices can do the organising, the islands are there at all².

They are large islands. Herbert Simon imagined in 1991 a visitor from Mars with a telescope that shows firms as solid green areas and market transactions as red lines between them. Its message home, he wrote, would speak of “large green areas interconnected by red lines”, and the visitor would be surprised to hear the scene called a market economy⁹.

A cost of using the price mechanism

Coase’s answer takes one sentence: “The main reason why it is profitable to establish a firm would seem to be that there is a cost of using the price mechanism.” Somebody has to find out what the prices are. Every exchange needs a contract, which has to be negotiated and concluded. Inside a firm one contract replaces many: a worker agrees, for a wage, “to obey the directions of an entrepreneur within certain limits”, and within those limits is told what to do. Coase sums up that “the operation of a market costs something” and that an organisation saves part of the cost².

The objection is obvious, and he raises it himself. If organising saves cost, “Why is not all production carried on by one big firm?” His reply is that organising has costs too, and that they rise as a firm grows, because whoever is in charge has more to keep track of and makes more mistakes. He calls this “decreasing returns to the entrepreneur function”. The two costs meet at a margin. A firm will expand, he writes, “until the costs of organising an extra transaction within the firm become equal to the costs of carrying out the same transaction by means of an exchange on the open market or the costs of organising in another firm”².

The core of the theory has no more parts than that. It turns the size of a firm into a matter of comparison, and in a footnote it gives Coase a reply to the planners and to their opponents at once. State planning is imposed on industry, he notes, while firms “arise voluntarily because they represent a more efficient method of organising production”. The footnote ends with an exclamation mark: “In a competitive system, there is an ‘optimum’ amount of planning!”²

The paper offers almost no evidence. A Lancashire weaver, a department store and a Russian factory appear in passing, but there is no table in it and no firm is studied. Its best-known term is absent: the words “transaction cost” appear nowhere, and Coase writes of “marketing costs”, eight times, and of the “cost of using the price mechanism” instead (own count). And it ends modestly. Its closing words describe the whole exercise as “our comparatively simple task of definition and clarification”². Rereading it half a century later, Coase was struck by “its extreme simplicity”¹⁰.

Congratulated on the way to lunch

He had a draft by the early summer of 1934, at 23, and published it with very few changes in November 1937, by which time he was teaching at the London School of Economics¹⁰. “On the day that the issue of Economica was published,” he recalled, “on the way to lunch I was congratulated by Sargent and Plant, the professors of commerce. Neither ever referred to the article again or discussed it with me”. The head of his department, Lionel Robbins, never mentioned it to him, and neither did Friedrich Hayek¹⁰.

The paper was not forgotten. George Stigler chose it for the American Economic Association’s volume of readings in price theory, published in 1952, which, Coase said, “increased the number of footnote references but did not bring about any real interest in the questions raised”¹¹. That is what he meant in 1970 by “much cited and little used”, and he gave a reason: the problems the theory was meant to illuminate “have not been of much interest to economists in recent years”⁵. His later article on social cost, by contrast, was in his own words “an instant success”³. The Swedish Academy’s summary in 1991 was that it “took a long time for his approach to gain a foothold”¹².

Cited most in the 1990s and 2000s

Of every 1,000 indexed works on a subject, the number that cite “The Nature of the Firm”, by decade of publication.

Show the numbers

Published	Economics: citing works	of all works	per 1,000	Business: citing works	of all works	per 1,000
1970s	40	29,838	1.34	37	18,044	2.05
1980s	246	68,812	3.57	172	39,890	4.31
1990s	658	127,625	5.16	836	100,937	8.28
2000s	1,589	354,124	4.49	3,263	339,190	9.62
2010s	1,962	678,416	2.89	4,483	878,774	5.10
2020–25	905	540,229	1.68	1,763	852,625	2.07

The record of citations fits that account as far as it can be checked. OpenAlex, an open index of scholarly work, sorts works by subject and lists the ones that cite the paper. For the years before 1970 it holds too few reference lists to be of use. Of every 1,000 works on economics, 1.3 cited the paper in the 1970s, 3.6 in the 1980s and 5.2 in the 1990s, the decade of the prize. The rate has since fallen to 1.7, although the number of works citing the paper went on rising, to nearly 2,000 in the 2010s: the index grew faster. In work on business and management the rate peaked later and higher, at 9.6 in the 2000s. By the index’s sorting, 45 per cent of all the works that cite the paper are about business and management and 23 per cent about economics (own count)¹³.

In the 1970s the paper acquired critics, and one economist, Oliver Williamson, set about making it testable.

A proposition which is clearly right

The sharpest early criticism came in 1972, from Armen Alchian and Harold Demsetz. Coase’s firm rests on authority: inside it, people do as they are told. Alchian and Demsetz called that a delusion. A firm, they wrote, has “no power of fiat, no authority, no disciplinary action any different in the slightest degree from ordinary market contracting between any two people”. An employer can fire or sue, which is what a customer can do to a shopkeeper. “Telling an employee to type this letter rather than to file that document is like my telling a grocer to sell me this brand of tuna rather than that brand of bread”¹⁴.

Their second objection went deeper. That firms do more where markets cost more to use, they wrote, is “a difficult proposition to disagree with or to refute”. One could “with equal ease” build a theory of the firm on the cost of managing: the lower it is, the more firms will do. Both statements are true, and neither says when a firm will take a transaction inside¹⁴.

The charge stuck, even among admirers. Six years later Benjamin Klein, Robert Crawford and Alchian, in a paper that set out to build on Coase, described his starting point as “the truism that profit-maximizing firms will undertake those activities that they find cheaper to administer internally than to purchase in the market”¹⁵. Williamson quoted the complaint of the economist Stanley Fischer that transaction costs “have a well-deserved bad name as a theoretical device”, partly, Fischer said, because of a suspicion that almost anything can be rationalised by invoking suitable ones¹⁶. In 2009 the committee that awarded Williamson his own Nobel prize wrote that the 1937 paper “offered only the preliminaries of an actual theory of the firm”. Without saying what determines the two costs, “Coase’s statement has little empirical content”¹⁷.

Theorists added two objections. The paper does not show why the costs should be lower inside a firm. Sanford Grossman and Oliver Hart put the point in 1986 against the whole family of transaction-cost arguments: they do not explain how the room for one side to exploit the other changes “when one of the self-interested owners becomes an equally self-interested employee of the other owner”. And they found Coase’s limit on the size of a firm, the owner’s limited capacity to manage, “unconvincing since the owner could always hire another manager”¹⁸.

Coase’s replies varied with the charge. Being told that his proposition was a tautology, true by definition, did not trouble him. “It is the criticism people make of a proposition which is clearly right,” he said in 1987, and Grossman and Hart, he said, had misread him¹⁰. Elsewhere he gave way. He named a weakness of his own: “I consider that one of the main weaknesses of my article stems from the use of the employer-employee relationship as the archetype of the firm”. On what decides between firm and market he conceded the point entirely. The paper contains “no such comprehensive theory”, he said in the same lectures¹¹. In Stockholm he repeated Williamson’s judgement that he “did not indicate what the factors were that determined the outcome of this choice”, and added: “This also is true”⁷. And in 2009, by the University of Chicago Law School’s account, he disputed the paper’s claim that returns diminish as a firm grows larger, and called the growth of firms a sociological matter⁶.

The part he left out

A theory that does not say what its costs depend on cannot be tested. Williamson supplied the missing part from 1971. Transactions differ, he argued, and the ones that belong inside a firm are those that are complex and that need investments of use to one trading partner only. A press tool that fits one carmaker’s door is an example. Economists call this asset specificity. The danger, described by Klein, Crawford and Alchian in 1978, is that once such an investment has been made the other side can reopen the bargain, and that owning the supplier, which is called vertical integration, may be the cheapest protection¹⁵. Economists came to call this the hold-up problem¹⁹. Because complexity and specificity can be measured, the Nobel committee wrote, “Williamson’s theory had the required empirical content”¹⁷.

Coase had met this argument in 1932 and had written it down himself. In a letter from Detroit that spring he described a supplier that must buy costly equipment “specialized insofar that it can only be used for the particular product concerned”, and that then faces “one great risk”: its only customer may take his business elsewhere or “exercise his monopoly power to force down the price”. Looking back in 1987, he said he had thought the argument sound and had come across cases of it. He had also found that firms wrote contracts to get round the danger and commonly made such investments for a single customer all the same¹. His example was A. O. Smith of Milwaukee, an independent maker of car frames whose main customer was General Motors¹¹. So the investment alone could not decide between a contract and ownership, and he went on looking for something more general¹.

The textbook case of hold-up is General Motors and Fisher Body. As Klein, Crawford and Alchian told it, GM signed a ten-year contract in 1919 to buy its closed car bodies from Fisher at cost plus 17.6 per cent. Demand then shifted sharply to closed cars, GM came to think the price too high, and Fisher, they wrote, “refused to locate their body plants adjacent to General Motors assembly plants”. By 1924 GM found the arrangement intolerable, and in 1926, already the owner of 60 per cent of Fisher’s shares, it bought the rest¹⁵.

In 1987 Coase’s own recollection of the case was still close to theirs. He had seen a Fisher Body plant in 1932, and remembered being told at General Motors that the main reason for the purchase “was to make sure that the body plants were located near General Motors assembly plants”. As things had looked to him then, he said, the case “might well have been an example” of ownership forced by this kind of risk¹. Thirteen years later he disputed it outright. In an article published when he was 89 he wrote: “This tale is factually incorrect.” Fisher had not put its plants far from GM’s, and “There is no evidence that a holdup occurred”²⁰.

Klein replied in the same issue of the journal and conceded a good deal. Coase’s conclusion “that Fisher generally located its plants close to GM production facilities during this early period, is accurate”, he wrote, and the 1978 paper had been mistaken to date the breakdown to 1924. What remained, for Klein, was 1925, when Fisher would not build a plant at Flint and bodies were hauled 57 miles from Detroit with 17.6 per cent added to the cost of hauling them. He still called that a hold-up and the main reason for the purchase²¹. Two other articles in the issue took Coase’s side. One, by Robert Freeland, argued that the Fisher brothers held General Motors up only after they had become its employees²².

One case settles nothing about a theory. It does show where Coase stood. “Coase rejects this idea that specific investments are a determinant of vertical integration,” Klein wrote²¹. The mechanism that made the paper testable was one he had weighed at 21 and set aside, and its showpiece was a story he ended by denying.

Overwhelming, mixed and modest

By the 1990s the testable version had been tested many times, usually by asking whether firms make for themselves the parts and services that need specific investment and buy the ones that do not. Reviews of this work reach verdicts that sound incompatible. Francine Lafontaine and Margaret Slade, surveying it for the Journal of Economic Literature in 2007, wrote: “The weight of the evidence is overwhelming”²³. A meta-analysis in a management journal the year before had reported strong support²⁴. Robert David and Shin-Kap Han, counting results for another management journal in 2004, found support “at 47 percent” and said they were surprised²⁵. A meta-analysis of 143 studies in 2013 credited the theory with “a real yet modest ability to explain how firms organize economic activity”²⁶.

The verdicts are less far apart than they sound, because each rests on a different question.

Under half of the tests came out as predicted

How 308 statistical tests of transaction cost economics came out, in 63 articles published from 1982 to 2002.

Show the numbers

Tested	Tests	Significant, as predicted	Not significant	Significant, the wrong way
All tests	308	144 (47%)	133 (43%)	31 (10%)
Asset specificity	107	64 (60%)	39 (36%)	4 (4%)
Uncertainty	87	21 (24%)	52 (60%)	14 (16%)

David and Han counted 308 statistical tests in 63 articles. A test counted as support only if it was statistically significant in the predicted direction. On that standard 144 passed, 31 were significant in the wrong direction, and 133 were not statistically significant (the last figure is our subtraction). The theory’s central variable did better: of 107 tests of asset specificity, 64 supported it and 4 contradicted it. Uncertainty, which is also supposed to push transactions inside the firm, was supported in 21 of 87 tests and contradicted in 14²⁵.

Lafontaine and Slade asked of a smaller set of studies, mostly from economics journals, whether anything significant contradicts the theory. Almost nothing does. In their eight tables on transaction costs we count 41 estimates from 22 studies, some of which appear in several tables: 35 point the predicted way, 30 of them significantly, and two are significantly negative (own count from the working-paper version)²³. The meta-analysis of 2013 asked how strong the link is. Across the 109 of its studies that measure it, the corrected correlation between asset specificity and integration is 0.12. Squared, that is about 1.4 per cent of the variation in how far firms integrate (own calculation), and the authors’ own phrase is that transaction costs “explain only a fraction” of what managers do²⁶.

So the reviews do not contradict one another. For the central variable the predicted effect appears in most tests and seldom reverses. How often it is statistically significant depends on the sample, from six in ten of the tests in the management literature to three-quarters of the estimates in the economists’ tables, and the one review that measured its strength found it weak. For uncertainty the record is close to a draw. Whether that amounts to overwhelming support depends on what one was hoping for.

The reviews list doubts of their own. Lafontaine and Slade note that specificity is mostly measured by asking managers about it²³. Another review points out that it is treated as given, although firms choose how specialised their investments are²⁷. The classic studies did not test a rival explanation, such as the view that firms keep in-house what they are good at: of 27 widely cited studies examined in 2006 none did, though the examiners name four later ones that do²⁸. There are plain counter-examples too. Bengt Holmström and John Roberts pointed out in 1998 that Japanese carmakers leave the specialised dies with their suppliers, where the theory expects the carmaker to own them, and that the relationships are stable all the same: only three of roughly 150 first-level Toyota suppliers had left its association in eleven years¹⁹.

A study of 2014 raised a different doubt. Enghin Atalay, Ali Hortaçsu and Chad Syverson used the Census Bureau’s shipment surveys of 1993 and 1997 to see where plants owned by vertically integrated American firms send their output. The textbook picture is that a firm owns its supplier to secure the part. They found that “roughly one-half of upstream establishments report no shipments to downstream establishments within the same firm”, and that the median plant sends 0.4 per cent of its shipments there. On two looser definitions of an internal shipment the share that sends nothing falls to about a quarter. Their reading is that ownership may mainly serve to move intangible inputs, such as know-how, within the firm, in which case the decision to make or buy a physical part, the decision the textbook cases are about, explains only a fraction of what firms own²⁹. The same authors later found that ownership does pull goods towards it: a plant of the same firm at the destination raises shipments by as much as a 40 per cent cut in distance³⁰. Multinationals resemble the plants in the first study: in 2004 the median foreign manufacturing affiliate of an American company shipped nothing to its parent and received nothing from it³¹.

Technology and the size of firms

The idea reached a wider public as a forecast. If firms exist because markets are costly to use, and computers make markets cheap to use, then firms should shrink. Thomas Malone, Joanne Yates and Robert Benjamin argued in 1987, building on Williamson and without citing Coase, that information technology “will lead to an overall shift toward proportionately more use of markets”³². In 1998 Malone and Robert Laubacher described an “e-lance economy” whose basic unit is “not the corporation but the individual”, and imagined a General Motors split into several dozen divisions that had outsourced most of what they did³³. In the same year Larry Downes and Chunka Mui attached Coase’s name to a Law of Diminishing Firms: “As transaction costs in the open market approach zero, so does the size of the firm”³⁴. When the gig economy arrived, The Economist explained it under the heading “Coase and effect”³⁵, and in 2017 a brief on the paper said the boundary between firm and market “might appear to be dissolving altogether”, though it added that firms were unlikely to wither away³⁶.

Coase had written something else. The paper’s one passage on technology says: “Changes like the telephone and the telegraph which tend to reduce the cost of organising spatially will tend to increase the size of the firm.” A footnote on the same page adds a condition. Most inventions change both the cost of organising and the cost of using the market, and “whether the invention tends to make firms larger or smaller will depend on the relative effect on these two sets of costs”. The telephone itself, the footnote says, would make firms smaller if it cut the cost of using the market by more². So the paper does not predict that firms will shrink. It names the condition under which they would. Asked in 1997 whether he was excited by the new electronic ways of doing business, Coase answered: “Yes, because I don’t understand them”³⁷.

The largest firms did not dissolve

Share of all employees of American private-sector firms who work for the largest and for the smallest firms, 1978 to 2023.

Show the numbers

Year	10,000 or more employees	Fewer than 20 employees
1978	25.0%	22.7%
1979	25.1%	22.1%
1980	25.6%	21.6%
1981	25.8%	21.7%
1982	25.1%	21.7%
1983	24.7%	22.4%
1984	24.1%	22.6%
1985	24.3%	22.2%
1986	24.2%	22.1%
1987	23.4%	22.0%
1988	23.3%	21.9%
1989	23.2%	21.5%
1990	23.6%	21.3%
1991	24.2%	21.2%
1992	23.8%	21.3%
1993	23.7%	21.4%
1994	23.6%	21.1%
1995	23.8%	20.8%
1996	24.5%	20.6%
1997	25.0%	20.3%
1998	25.5%	19.7%
1999	26.1%	19.4%
2000	26.7%	19.0%
2001	26.8%	18.8%
2002	26.7%	19.3%
2003	26.2%	19.3%
2004	25.9%	19.4%
2005	26.2%	19.2%
2006	26.3%	19.1%
2007	26.7%	18.8%
2008	26.8%	18.5%
2009	27.2%	18.6%
2010	27.1%	18.9%
2011	27.3%	18.6%
2012	27.3%	18.4%
2013	27.8%	18.1%
2014	27.7%	17.8%
2015	28.1%	17.6%
2016	28.3%	17.4%
2017	28.5%	17.2%
2018	28.9%	17.1%
2019	29.2%	16.8%
2020	29.4%	16.6%
2021	30.0%	17.2%
2022	29.5%	17.3%
2023	29.9%	16.7%

What happened to firms can be read from the Census Bureau’s Business Dynamics Statistics, which cover American private-sector firms with employees. In 1978 firms with 10,000 or more employees accounted for 25.0 per cent of the employment in the series. Their share fell to 23.2 per cent in 1989, the lowest of the series, moved little until the mid-1990s and then rose, to 29.9 per cent in 2023. Firms with fewer than 20 employees accounted for 22.7 per cent in 1978 and for 16.7 per cent in 2023 (own calculation)³⁸.

The forecast rested on something real. Erik Brynjolfsson, Malone and two colleagues found that investment in information technology was followed by smaller firms, in data for 1976 to 1989, the years in which the largest firms’ share was falling³⁹. Malone and Laubacher’s evidence in 1998 was that the share of American workers employed by Fortune 500 companies had fallen from one in five to less than one in ten in 25 years³³. Until 1995 the Fortune 500 list ranked industrial companies only⁴⁰, and in manufacturing the trend they saw went on. A Census Bureau study of the same series reports that manufacturing kept moving towards smaller firms while retailing moved the other way, the share of its employment in large firms rising from 36 per cent in 1978 to 62 per cent in 2019⁴¹. Across the whole economy the largest firms’ share was back above its level of 1978 by 1998, the year the e-lance economy was announced.

So the paper does not predict that firms will shrink.

Workers did not leave for self-employment in the numbers predicted either. The Bureau of Labor Statistics, counting people by their main job, found independent contractors to be 7.4 per cent of the employed in February 2005, 6.9 per cent in May 2017 and 7.4 per cent in July 2023⁴²⁴³. In Germany people who work for themselves without staff were 3.6 per cent of the employed in 1991, 6.1 per cent at the peak in 2005 and 3.5 per cent in 2025. The statistical office links the growth from 2003 to a start-up subsidy, the so-called Ich-AG, which was replaced in mid-2006 by a narrower one⁴⁴. One survey seemed to show a boom. Lawrence Katz and Alan Krueger reported in 2016 that alternative work arrangements, a wider category that also takes in agency, on-call and contract-company workers, had risen from 10.7 to 15.8 per cent of American workers in a decade⁴⁵. Three years later they revised this to “a 1 to 2 percentage point increase” and wrote of “our 2016 overstatement”⁴⁶.

The forecast was right about some things. Firms did hand work to the market, though work of particular kinds. In West Germany the share of the workforce employed by cleaning, security and logistics contractors and by temporary agencies rose from 2 per cent in the mid-1970s to almost 8 per cent in 2008, most of it after 1990. About one cleaner in ten worked for a cleaning firm in 1975, and almost four in ten in 2008⁴⁷. In the United States, by a wider definition and over a longer period, the share of the workforce in business service firms rose from under 2 per cent in 1950 to more than 8 per cent in 2015⁴⁸. A study that followed individual firms points the same way: French firms that got broadband in the 2000s spent more on outsourcing and employed a narrower range of occupations⁴⁹.

Other measures are harder to read. The number of tax returns filed by American businesses with no employees doubled, from 15.4 million for 1997 to 32.0 million for 2024 (own count from Census files)⁵⁰. How much of that is new work is disputed: a study of the tax records published in 2025 finds that the rise in reported self-employment since 2005 is not matched by payments that firms report making to contractors, and comes mainly from what low earners declare⁵¹. By one count the largest firms have narrowed their range, the biggest firm in an average industry being active in 13 other industries in 1982 and in fewer than 9 in 2012⁵². A study of listed companies’ annual reports finds the opposite, that American firms have widened what they do over the past thirty years⁵³.

The number of people a firm employs is in any case a rough stand-in for what the paper meant by size, which was the number of transactions a firm organises. No series counts that. The nearest are the share of America’s foreign trade in goods that takes place between related companies, 38.5 per cent in 1992 and 42.8 per cent in 2025 (own calculation for 1992)⁵⁴, and the share of corporate assets held by the largest 1 per cent of American corporations, which rose from 70 per cent in 1931 to 97 per cent in 2018⁵⁵.

One study shows both of Coase’s costs at work in a single industry. George Baker and Thomas Hubbard studied what on-board computers did to American haulage between 1987 and 1997. Simple trip recorders, which let an owner see afterwards how a driver had handled a truck, made it cheaper to supervise one’s own drivers and pushed haulage into shippers’ own fleets. The more advanced devices also reported a truck’s position as it travelled, which helps a dispatcher find return loads, and that capability pushed haulage out to trucking companies. “Has IT adoption led to larger or smaller firms in trucking?” the authors asked, and answered: yes. They left open whether the result holds in other industries⁵⁶.

What is left

Three things have to be kept apart: the question, the answer and the forecast.

The question has lasted. Even the critics who denied that a firm runs on authority went on to ask why firms exist and where they end, and that is now a field with prizes of its own: Williamson’s in 2009 for work on the boundaries of the firm¹⁷, and in 2016 that of Hart and Holmström for contract theory, whose committee named the 1937 paper as “the intellectual origin” of the modern theory of those boundaries⁵⁷.

The answer, as written, was a way of looking. It said where to look, at the cost of transacting set against the cost of organising, and did not say what would be found. Coase agreed, and said in Stockholm that what was needed was “more empirical work” on contracts and on what firms do⁷. The version that could be tested was built by others, on a mechanism he had set aside. The tests bear out its main prediction, weakly, and its other predictions unevenly.

The forecast was never his. He wrote that the telephone and the telegraph would tend to make firms bigger, and added in a footnote that the telephone would make them smaller if it cut the cost of using the market by more. That cannot be wrong, which is the weakness his critics found in the paper as a whole: a statement that firms will grow or shrink according to which cost falls further fits any outcome. Its merit is that it says what to measure. Where somebody did, in trucking, the answer came out in both directions. And the pattern of the past thirty years, a growing share of jobs in the largest firms together with more outsourcing, is what the Law of Diminishing Firms did not expect.

Coase died in Chicago on 2 September 2013, at 102⁶. In Stockholm he had told his audience: “I have made no innovations in high theory.” His contribution, he said, had been to urge economists to take account of “features of the economic system so obvious” that they had been overlooked, and he found it “a strange experience to be praised in my eighties for work I did in my twenties”. He ended the lecture with the consolation open to a scholar whose idea is true: he “can count on ultimately seeing it accepted, if only he lives long enough”⁷.