FAIR USE NOTICE. This document contains copyrighted material whose use has not been specifically authorized by the copyright owner. The Managerial Economics course is making this material available as part of our mission to promote critical thinking about economic issues. We believe that this constitutes a `fair use' of the copyrighted material as provided for in section 107 of the US Copyright Law. If you wish to use this copyrighted material for purposes of your own that go beyond `fair use', you must obtain permission from the copyright owner.

The New Australian Magic puddings, Microsoft and the Myth of Increasing Returns in Software

By Gerard Jackson No. 78, 8-14 June 1998

The nature of the software industry has caused some economic observers to claim that economies of scale have given companies like Microsoft an unassailable lead that will result in monopoly. For that reason governments must act to restrain the growth of these companies in order to prevent them from crushing competition. These observers rest their case on the assumption that once software has been produced marginal costs (the additional cost of producing one more unit) fall to virtually zero. These low production costs combined with very high development costs create, according to this view, an insurmountable barrier to potential competitors thus leading to industry concentration. It is, however, the contention of this article that these observers are gravely mistaken and that in fact economies of scale do not prevail in the software industry. Moreover, high development costs are not acting as a barrier to competition. As this article is solely concerned with the argument that economies of scale are the main barrier to competition, the supposed role of development costs in promoting monopoly will be dealt with in another article.

Before continuing, it is necessary to first explain what economists mean by economies of scale. This condition prevails when additional physical inputs result in a disproportionate increase in output which causes the firm's long run average costs of production to fall. (Falling average costs always exceed marginal costs.) Now the real key to economies of scale is indivisibility. There is a minimum size for the optimum use of a factor or group of factors. If the optimum is too large for a small or medium sized plant than a larger plant must be built. It is self-evident that indivisibility is the reason for the larger plant. A blast furnace is a graphic example of an indivisibility. A single blast furnace obviously yields greater efficiencies than 40 or 50 backyard furnaces. Equally obvious is the fact that we cannot subdivide a blast furnace, i.e., it is not divisible.

Professor Stigler gave a railway track as an example of an indivisibility and its influence on output (Theory of Price). The optimum output for the track is 200 trains a day. Any deviation from optimum will lead to a fall in output. Less than 200 means that the divisible factors, mainly staff, will have to be reduced while the indivisible factor, the track, remains unchanged and underutilised. If more than 200 trains a day are run, the quantity of divisible factors will have to be increased and the track will be over-utilised because output now exceeds the track's technological size which is set for 200. Therefore raising output to 200 generates increasing returns (falling average costs) while any increase in excess of 200 will generate decreasing returns (increasing average costs). From this we can easily deduce that when an important indivisible factor is being overutilised the tendency will be for increasing average costs, and decreasing average costs if the factor is becoming more fully utilised.

In reality it is not always so straightforward. Manufacturers usually find themselves in a situation where the factors differ in size and divisibility. These factors have to be combined in way that at least attempts to achieve an optimum outcome. This frequently means that the firm finds itself in a the situation where it can fully utilise one indivisible factor but not the complementary factors. At this point costs of production will rise. Bringing in another indivisible factor to fully utilise the complementary factors will see average costs fall. Where economies of scale are being achieved these economies will be further increased where they have resulted in greater specialisation among the complementary factors, including services and management.

The reader is now aware of the pivotal role of indivisibility in any discussion of economies of scale. Yet indivisibility is the very thing that seems to be missing from the debate on the alleged role of economies of scale in the production of software. Readers will also note that economies of scale involve a continuous production process in which increasing output leads to a fall in unit costs. What appears to have been completely overlooked is that software production only involves a single unit and thus precludes economies of scale.

Any firm enjoying economies of scale increases its net revenue by expanding sales through lower prices made possible by falling average costs. The more units it sells the greater its net revenue will be so long as average costs fall, despite the fact that total costs rise with output. However, the main production costs for a software company is in development of the product. Once the product is developed, the development team can be literally dispensed with leaving only the software package. This brings us to the nature of software. All software products are identical in the sense that they consist of instructions, lines of code or technological recipes, if you like. And like recipes or formulaes their services are inexhaustible. Only the existence of copyright (with the exception of freeware) prevents them from becoming free goods. The vital point is that once they have been produced there need virtually be no further production costs, meaning that marginal costs will be zero.

For example, a group of progammers develop a software package (let's call it SpeedNet), post its existence through news groups telling them its freeware. As we can see, there are no longer any production costs. The only costs were in development and they have now passed away. Bygones are always bygones. Let us also assume that this software becomes phenomenally popular, as did Doom, with hundreds of thousands of net users downloading it every week. Does this mean that economies of scale prevail? Of course not. What these users are doing does not differ in principle from millions of students copying out Pythagoras' theorem. Downloading, therefore, is a form of copying. (This, I believe, is an extremely important point.) Staggered by their success our programming enthusiasts develop a compression algorithm that is superior to anything on the market. Keen to make money, they now charge a modest downloading fee. The programme is so successful it now yields our budding entrepreneurs several million dollars a month.

That a downloading fee is charged obviously does not alter the fact that no further costs of production are incurred, no matter how many times the programme is downloaded. Though it is true that our heroes now incur administrative costs, these costs will tend to be 'fixed' rather than variable. (See Costs, markets and economic reform for a detailed view of costs). Burning with ambition our young entrepreneurs use their newly acquired wealth to form a company to continually upgrade their compression programme and produce other programs. Offices are rented and technical teams formed. What started as a game has now become big business. Nevertheless, the economics remains basically the same.

It can be argued that with more complex programs requiring ever bigger teams of programmers, these teams become indivisible factors. Quite so. But that does not change the fundamental fact these factor combinations are being used to produce a single product and not a continuous flow as in a production line. Once, for example, Ford's development team have designed a car, production lines are retooled and economies of scale then permits mass production on a continuous basis. But if by some technological miracle consumers could download Ford's latest model then Ford would close down its factories and economies of scale would no longer have any role to play in the company's operations. It would be in the same position as a software company.

We do not have to look to software companies to find the phenomenon of a single good providing a continuous service at zero marginal costs and at negligible average costs. Lighthouse services provide a graphic example of this kind of good (See R. H. Coase, The Lighthouse in Economics). Once the lighthouse has been built there only remains operating costs, which in turn are completely unrelated to the number of ships that use the service. This means that no matter how many ships use the service there are no additional costs, i.e., marginal costs are zero. Yet no one has claimed — at least to my knowledge — that any lighthouse operations have been subject to increasing returns. Another example would be that of charging to look at an art exhibit. As there no extra costs are incurred in allowing additional patrons to view the exhibit does this mean increasing returns for the exhibitor? Of course not, and no economists would ever say otherwise. Why? Because if average costs fall marginal costs must also fall. If marginal costs are zero then average costs must be constant or non-existent. As the reader must now know, It is falling long run average costs that defines increasing returns. If these costs are missing, so are the economies of scale.

The example of the lighthouse and the art exhibition should make it clear that these goods are no different from software in the sense that once they have been produced their services involve no additional production costs, regardless of the level of total consumption. In fact, the industry may well have served us a magic pudding. No matter how much of the services of its products we consume, the products remains inexhaustible.

I think it is now clear that the confusion about software and increasing returns springs from the erroneous assumption that the extremely low production costs (zero in reality) of any software product's services must have been the result of economies of scale driving down costs. This error derives from the failure to distinguish between a single good that can virtually provide an infinite flow of services to any sized market and goods, usually physical, that have to be produced as individual units for their services to be consumed, e.g., cars, televisions, bread, furniture, etc. Put another way, when consumers buy software they are really buying access to the services of the same programme just as someone who buys a season ticket to his football club's enclosure is buying access that is shared by other ticket holders. The basic difference being that a football club's enclosure has physical limitations.

Whatever barriers of entry might exist in the software industry, economies of scale is not one of them.

Note:

I have left out the costs of marketing and packaging because I believe these have little if any bearing on the view that increasing returns give companies like Microsoft monopoly powers.

The New Australian