post-autistic economics review
Issue no. 39, 1 October 2006
article 3



      issue 39 contents                            PAE Review index                               home page



Will Computers Really Decentralize the Economy?

Ian Fletcher   (USA)



            It is a ubiquitous assumption of the technorati that future advances in computer technology will further decentralize advanced economies.  But this assumption is probably autistic, and the reality a lot more complicated.



Basic Analytical Categories


            It is an error to view computers as having the same effect on every industry or part of society.  Instead, one should divide the world into four categories:


            1. Areas where computers are a centralizing force.


            2. Areas where computers are a decentralizing force.


            3. Areas where computers cut both ways.


            4. Areas where they have no effect.


The overall effect of computers on business and society is an average of these four effects. 


            These effects change with the state of computer technology.  Prior to about 1975, the relatively primitive state of computers favored category #1, as computers were themselves centrally-controlled mainframes, and were so expensive that only large centralized organizations could afford them.  Post-1975,  computers themselves physically decentralized and became cheap, so this factor gradually dissipated.  



Basic Cause-and-Effect Model


            In 2006, corporate America’s use of computers is dominated by the drive to exploit the economies of scale inherent in scalability.  Although technological breakthroughs, which enable computers to perform tasks they could not previously perform at all, capture the public’s imagination, most of these breakthroughs are economically (as opposed to technologically) viable for the same reason as most productive technologies since the dawn of time: a large quantity of work can be funneled through a finite technology infrastructure.  If this were not so, then purely technological advances, like the ability of computers to recognize human speech, would remain like the moon landings: technological feats of trivial economic significance.  But where will this scalability trend lead, as no trend lasts forever in economic history – something commentators often forget.  


            The basic assumption we should make is that as computers become ever cheaper, more capable, and more familiar, getting them to do whatever is called for by the economic task at hand will become ever more trivial.  As a result, computer technology will cease to be felt as a constraint on what can or should be done; it will become a “free variable” that can be effortlessly organized around other needs.  As a result, the key trend will be this:


As technology becomes more liquid, technological factors will cease to dominate organizations.


It follows that:


As technology becomes more liquid, non-technological factors will increasingly dominate organizations.


Superficially, this is paradoxical, but it makes perfect sense in terms of economic history.  For most of human history, agriculture was primitive, so producing enough food to feed the population was a major challenge.  A huge percentage of the population was engaged in farming, and major social institutions were organized around it.  But as agriculture became sophisticated, food supply became something that could be taken for granted, and food production today occupies only a tiny percentage of society’s attention.


            This does not mean that computers will become unimportant, any more than food has.  But it does mean that corporations, for example, will increasingly be organized around non-technological factors, like purely economic economies of scale, and decreasingly around their ability to afford and operate computers.  (I am not saying the state of computer technology has been the only determinant of corporate organization; only that it has been one factor, which will decrease in importance.)


            Take the giant American retailer Wal-Mart.   Today, it enjoys a competitive advantage because of its superb computerized supply-chain management, which gives it sophisticated control of its supply chain at a low cost per item sold.   But if any company could cheaply buy a similar system off the shelf, this would cease to be a competitive advantage, and thus remove a major factor that presently favors the existence of this highly-centralized company. It is only because Wal-Mart’s logistical rocket science is expensive, that it takes a multi-billion-dollar company to spread that cost over enough items to make the cost-per-item viable.  If it were cheap, a small company could afford logistics as good as Wal-Mart has.


            But this does not mean Wal-Mart will go away if this technology becomes cheap, as this is not the only competitive advantage the present centralization of the company generates.   It also produces a number of economies of scale, starting with the logistical advantages of being big that exist even without computerized management thereof.  The fact that a small competitor could one day cheaply copy its computer systems will not give that competitor its fleet of trucks and distribution centers. Nor will it give its competitor Wal-Mart’s buying power, or change the fact that investors would rather put their capital into a well-understood company like Wal-Mart, than into Jack’s 99-Cent Emporium of Hoboken, NJ.


            Economies of scale are not the only non-technological factors that will increasingly “show above the water” as cheap technology renders technological factors moot as a source of competitive advantage.  For example, the New York Times used to have a uniquely privileged position as a distributor of news to educated and affluent Americans, simply because it had the physical means to dump millions of newspapers on doorsteps nationwide every morning.  Small start-up publishers simply could not break into its market.


            But with the Internet, and mere ownership of printing presses and delivery trucks no longer conferred privileged access to these readers.   This story is familiar.  But the reason the Times has not collapsed, of course, is the fact that these were not its only competitive assets.  It also has, in addition to intangibles like brand image, a huge stable of capable writers and editors, who are able to turn out a culturally-sophisticated product that few can duplicate.   This sophistication is most visible in the soft-news sections of the Sunday Times: anyone who has tried to run an Internet magazine, or who compares the Times’s lifestyle and culture sections with what gets printed in regional newspapers, will be forced to admit this is true.  In this lies the Times’s true remaining competitive advantage. 


            It will be similar in other industries: cheap computers will tend to boil away technology as a source of competitive advantage, raising the relative significance of other factors.  As a result, the future shape of any industry, (or branch of government or aspect of culture) with respect to centralization and decentralization, will increasingly depend on this: what do the non-technology factors favor? 


Paradoxically, computers will thus over time kill their own significance.



The Long-Term Picture


            The obvious implied question, in the case of the Times above, is whether “cultural sophistication” on its own is really a centralizing factor.   This is a very tricky question:  on the one hand, increasing diversity of cultural and lifestyle options argues for decentralization of the raw expertise that constitutes sophistication, and an explosion of sophisticated lifestyle publications confirms this fact.  The idea of a few hundred Manhattanites being the arbiters of educated American culture was perhaps plausible as late as 1988, but not today. But on the other hand, raw expertise, as any penniless cultural critic or bankrupt magazine entrepreneur will know, is not an economic commodity.  It has to be packaged into a form consumable by affluent consumers for it to be worth money. 


            It follows that this packaging would seem to be where the Times’s true competitive advantage lies, so whether packaging is a centralizing or decentralizing force becomes the key question.  A lot of this comes down to whether the management of the Times really knows something other companies don’t: do they have a method for producing and packaging cultural sophistication ad infinitum?  If they can, then they’ve got a unique (or rare) skill.  And anything unique and valuable is centralizing by definition.  So the key determinant of centralization here will be whether this kind of skill tends to crystalize in a small number of places, perhaps because it depends upon face-to-face interactions within small teams of people.


            But what about more mundane kinds of management?  Returning to the Wal-Mart case, let’s look at how the company will hypothetically look, a few years after its computer systems have become available to anyone. 


            For a start, it will remain easier for Wal-Mart to attract capital, than for any given non-chain store selling the same goods –  say, Jack’s 99-Cent Emporium.  This is, at bottom, because the cost of doing proper financial analysis on Wal-Mart, sufficient to know that the company is worth investing in, is no greater than doing the same analysis on a company half Wal-Mart’s size, not much greater than doing it on a company 1/10 Wal-Mart’s size, and not all that much greater than doing it on a company 1/100 Wal-Mart’s size.  So the economy of scale in attracting capital, is at bottom an economy of scale in financial analysis, and as long as financial analysis is both expensive and scalable, it will tend to favor centralization.


            But if financial analysis ever becomes cheap, economies of scale in access to capital will cease to be a centralizing force (unless, of course, other factors turn out to affect economies of scale in access to capital.)   Cheap financial analysis would probably require science-fiction levels of artificial intelligence, but is not impossible in the long run.  Early stages of this are already visible: for debt capital, computerized innovations like credit scoring have already drastically reduced the cost of financial analysis, and there is strong evidence that small firms pay a lower premium on bank loans since its introduction. 


            A similar dynamic is likely with the internal corporate management of companies like Wal-Mart.   Presently, there are some very expensive MBA’s in Bentonville, Arkansas running the company.   Because applying their decisions to 1,000 stores costs trivially more than applying them to 100, it is efficient to centralize stores under their management. But what if cheap robot MBA’s became available?  Then this would cease to operate a centralizing factor.  


We may generalize thus:


Technological advances at the present level of technology, like decreases in price and increases in ubiquity and ease, increase the significance of non-technological factors.  Technological advances to new levels of technology, changes these factors.


            Whether the latter advances will, in any particular case, be centralizing or decentralizing, will depend.  If McDonald’s can replace its MBA’s with cheap robots, this will weaken the economy of its management, because any company will be able to afford management of similar capabilities, and therefore be a decentralizing force.  But if McDonald’s can replace its counter clerks and hamburger cooks with robots, so that thousands of restaurants can be remotely managed from a single control room at headquarters, it will be centralizing. 


            Even absent technological encroachment upon the frontier between technological and non-technological factors, non-technological factors themselves evolve.   And increasingly liquid computer technology removes implementational “friction” from the economic environment, so that this evolution is more intensely reflected in economic structure. 


            For example, one of Wal-Mart’s non-technological centralizing factors is buyer power (in the sense this term is used by MBA’s, quintessentially in Michael Porter’s book Competitive Strategy, summarized here: .)  But basic economic theory tells us that buyer power only exists in markets that are not commodities:  no buyer can get cheap crude oil.


            As a result, over time, Wal-Mart may (or may not, if it evades commoditization of its goods by any of the known means) lose buyer power as a centralizing factor.  If our imaginary Jack’s 99-Cent Emporium can get scissors from Guangdong at the same price Wal-Mart can, decentralization will rear its head again.


            Let’s run a hypothetical scenario of the disaggregation of Wal-Mart, not because this will necessarily happen, but to identify the factors that make Wal-Mart be the way it is:


Stage 1: Wal-Mart as it is today.  Centralized distribution, same retail price everywhere for products like scissors,  same “wholesale” internal transfer price to every store. (Whether the company actually has these same prices today is irrelevant to laying out this thought experiment, which can accommodate any empirical particulars.)


Stage 2: Wal-Mart uses its computers to realize the market for scissors in Chicago and assigns a higher price than in Atlanta, so it raises the retail price, the transfer price, and/or the quantity supplied to the store.


Stage 3: Wal-Mart realizes that stage 2 is a bureaucratic response to price signals, not a market one, and replaces this with a simulated “internal market,” in which stores “bid” against each other for “wholesale” scissors from its supply chain.  (Internal markets of various kinds have been tried in a number of companies, like the Koch natural gas company.)


Stage 4: Wal-Mart jumps from a simulated free market to a real one, and breaks up the pieces of the company into independently-owned stores, distribution centers, etc, which freely contract for each other’s goods and services, rather than having them assigned by commands from headquarters.


            Although there are myriad issues at each stage here,  this thought-experiment makes clear that, in a sense, decentralization is the natural condition of economic life under a basically free market.  Therefore, if we see centralization instead, it is because some factor interfered with the transition between the stages above, which would otherwise run their course and decentralize everything. This is, of course, just basic Coase Theorem economics: firms exist at all, and big firms are bigger than small firms, because of transaction costs (broadly-defined) and the effectiveness of economies of scale in reducing these costs.  (See Oliver Williamson’s 1985 book The Economic Institutions of Capitalism, which applied the Coase theorem to the question of the boundaries of the firm.)  So the answer to the question of computers’ effect on centralization and decentralization is in the end obvious, almost trivial:


Computers centralize when they strengthen economies of scale.   They decentralize when they weaken economies of scale.


            This formulation covers the different possible outcomes we have seen or may see:


1. Computers did increase economies of scale in the past:  from about 1950 to1975; only large companies could afford them, making centralization a source of competitive advantage.


2. Computers may increase economies of scale in the future: in 2020, the robotic McDonald’s may enable McDonald’s to run the entire chain from a single control room at headquarters.


3. Computers did reduce economies of scale in the past: post-1994, the Internet made nationally-distributed media, and sales channels for non-media products, vastly easier to create.


4. Computers may reduce economies of scale in the future, if artificial intelligence weakens economies of scale in management by mass-producing cheap management skills.


            The above is a caricature: most effects of computers will be less dramatic.   But it nicely reveals the fundamental tension present: between computer advances that act like a robotic cook, and advances that act either like a robotic MBA or like the Internet.  The first will centralize, the latter two will decentralize. 


            So we have here one centralizing dynamic and two decentralizing dynamics.  Computer advances that make it easier to aggregate vast numbers of things, like retail purchases at Wal-Mart, are a centralizing force, when this aggregation does something economically useful, like enable the exploitation of an existing economy of scale, like Wal-Mart’s management or buyer power.


            But computer advances that destroy some economy of scale will be decentralizing, and interestingly, this can happen in two different ways.  With a robotic MBA, an expensive central resource suddenly becomes cheap, though the price of replicating its activity over the many objects of that activity remains the same.  With the Internet, the cost of the central resource (content) remains about the same, but the price of replicating that content to many consumers of it drops dramatically. 


            The above analysis implies that the deck is probably stacked in favor of decentralization – but only long-run and big picture.  In the short and medium term, and in particular industries, there exist profound reasons why things can cut the other way for long periods of time:


Until every existing economy of scale is liquidated (either by computers or something else) computers can often make it easier to exploit these economies of scale, and thus promote centralization.


            The above is only an analysis of economic factors.  Although political bureaucracies will to some extent be affected similarly to industries in how they implement policies, the fundamental fact of politics is not bureaucratic implementation, it is coercive power – which operates according to very different rules, completely outside the scope of the above discussion and under no obligation to behave the same way.  So as far as politics impinges on industry structure, the above analysis will be incomplete.




Ian Fletcher, “Will Computers Really Decentralize the Economy?”,
post-autistic economics review, issue  no. 39, 1 October 2006, article 3, pp. 26-30,