The Two Economies

Graham SeamanMail link

Or: Why the washing machine question is the wrong question

Note: There is also a transcription of the talkLocal link.


Within capitalism, material goods are typically made:

The co-ordination between producers is indirect, through the market, using money as a signalling mechanism.

Production of free software and other free goods can be contrasted point by point with this list; non-material goods can be produced by people:

The co-ordination between producers is direct, mediated only by technology.

In traditional marxist terms, two societies described like this would have different modes of production. But in this case there is only one society, and while almost the whole of society produces in the first way, only a tiny, though growing, part produces in the second.

This is not an unusual situation: there have been few times in history when a 'pure' mode of production, unmixed with fragments of other modes, existed. Some of these fragments are remnants of the past: personal slavery in parts of Northern Europe during the middle ages, or villages with communally allocated and rotated land in isolated parts of Southern Europe today. These fragments can often survive for long periods, integrated into the overall system and partially changed from their original form, but stable. Others are abortive glimpses of a future believed possible which turns out not to be so, such as the numerous experiments in communal working and living from the nineteenth century to the 60s and 70s of the last century, again often surviving for long periods. But the most interesting possibility is the fragment which turns out to be the replacement for the dominant mode of production.

This leads to two major groups of questions:

Firstly, what are the effects of the coexistence of two modes of production now? How does the dependence of free software producers on the capitalist economy affect free software production? And what effect, if any, does free software production have on the surrounding capitalist mode of production?

Secondly, is it possible for the free software mode of production to be generalised to the whole of society? And if so, how?

Obviously, these are questions without definitive answers. Even those parts of the question which are purely empirical would need a major research program to answer properly. But that doesn't mean that it is pointless to try to suggest possible answers. One possible starting point is to look to the past, to one of the best documented changes: the break-up of the feudal system in pre-revolutionary England.

The End of the Guilds

Manufacturing in late mediaeval society was contained within the guild system, and organised through the hierarchy of apprentices, journeymen and masters. To have a trade it was necessary to have been an apprentice; once apprenticeship had been completed (normally after 7 years) an apprentice could expect his master to register him as a full guild member, with the freedom to practise the trade as an independent journeyman. Naturally journeymen would expect to become masters in their turn. Knowledge of the trade was part of the mystery of the guild, shared vertically within the guild but kept a secret from outsiders, and guild boundaries were rigourously enforced. Guild inspectors would check not only the quality of the goods produced but also adherence to proper employment procedures and encroachment on the territory of other guilds: a shoemaker in the shoemakers guild should not encroach on the work of cobblers, who repaired old shoes, nor should he tan his own leather, the mystery of the tanners' guild. The system was intended to maintain the maximum possible quality of the output: the quality of tanned leather was guaranteed by the tanners' own inspectors, the true experts on tanning, and a shoemaker who set himself up as an amateur tanner as well had no such expertise.

By the late 16th century this system was still firmly in place. To some extent it was cross-cut by the patents of monopoly granted by the state, which effectively gave guild privileges to small groups of individuals (though even these were limited to 7 years, the time for a group of apprentices to pass through the system and potentially be able to set up a new guild); but the right of the state to grant such patents was fiercely (and often successfully) resisted by the guilds.

What the guilds could not do was cope with the increasing number of journeymen with no hope of becoming masters in their own guilds. In the big cities desperate journeymen began to abandon their own trades and set up as small manufacturers. These small manufacturers, though persecuted, managed to survive outside the guild system and the mediaeval hierarchy of rights and obligations, and in spite of the many caught by guild inspectors and fined or even imprisoned, by the mid-17th century parts of London were dominated by them. Since they were outside the guild system their employees were not apprentices in the old sense, but workers for a wage: this was already a fragment of a new mode of production.

Now two systems co-existed: one still dominant, the other small and struggling, and blocked at every turn by the regulations of the old system. John Lilburne, one of the leading spokesmen for the Levellers, the Republican left-wing, was a typical example: originally apprenticed as a clothier, he became a Protestant. Book publishing and distribution was a monopoly of the Stationers', and when he attempted to bring in Protestant texts from Holland he was caught by inspectors for the Stationers' Company and imprisoned. Once freed, he became a successful small brewer until the outbreak of the Civil War. After two and a half year's fighting, he attempted to use his knowledge of cloth as a cloth-exporter; but the monopoly on cloth export belonged to the Merchant Adventurers, not the clothiers themselves. Abandoning this, he became a soap-maker ... Just to survive, people like John Lilburne were forced to work outside, and against, the guilds.

Other Leveller supporters worked in brewing, tanning, glass-making, felt-making, hat-making, sack-cloth and linen-weaving, dyeing, silk-spinning, soap-boiling, nearly all embroiled in continual struggles with the guilds. It was natural that their watchword became 'freedom': freedom from the guilds, freedom from the state-imposed monopolies, freedom for trade, freedom of conscience.

So we have a first requirement: the new mode of production is not something arbitrary, willed into existence, but a product of the old system: in this case the guild system which was structurally unable to provide positions for all its apprentices.

Next, the new system began to infect the old. Here the route was simple: for the new mode of production to expand, it needed capital, and capital was already available. Merchant trading was a normal part of the mediaeval economy; once again, monopolised by merchant guilds. But given new possible sources of profit why should they care whether the products they traded had been produced under normal guild regulations or not? From reselling non-guild products it was a small step to financing their production, although in the end the restrictions on doing this on a large scale were too great, and the major new capitalist industries were not based on the original ones in the warrens of London, but in the North, away from any guild control at all. Once these large-scale industries had become established, the guild system was effectively doomed: the number of apprentices who could be integrated into the guild system with it's progression of stages was tiny compared with the mass of labourers required for the new manufactories. Some in the old system attempted to compete by taking on large numbers of apprentices against their own rules, or by employing journeymen who had not completed apprenticeships, but the result was that the guilds simply became empty ceremonial shells of their former selves, gradually to disappear over the next two centuries.

It is noticeable that the change from guild production to capitalist production was in its early stages not driven by technological change, but by the inability of the guild system to cope with expanding markets. The changes, and the casuses of the spread of the new system, were social. New technologies - in particular the use of steam-power in production - only became important a century later.

All this suggests some possible properties needed for a new mode of production to spread:

The statement that 'free software is the kernel of a new mode of production' often leads to the question 'how can you make washing-machines in the same way'? This depends on your assumptions about what that way consists of: is the primary fact technological, the fact that reduced costs for computers have made software effectively a public good; or is it social, and the fact that people are working together in a new way that is primary?

If it is the first, then production of material goods in the same way needs them to be 'dematerialized': we must wait for the invention of matter transmitters before it becomes possible.

If the second, then it is possible to give a more optimistic answer: once working by free software principles has spread far enough throughout the economy that it reaches the people who make washing machines, they will know how to do it. In every revolution of the last hundred years, people have begun to take control of their own work. If the revolution has been defeated, their control has been taken away. If the revolution has won, their control has been taken away. But the possibility is there, and has been shown repeatedly, even though it rarely appears in history books. What free software has proved that is new is the possibility of this style of work on a large scale, sustained over a long period of time.

But in either case, to expect a solution to the 'washing machine question' now would require magic; a sudden jump, whether technological or social, which is not likely to happen.

The Old Economy

The fact that software is part of the leading sector of the economy hardly needs justification, though in fact it turns out to be very difficult to show in any quantitive way; the obvious thing to do would appear to be to compare first the monetary value of so-called 'IP goods' compared with that of material goods, then the change over time of this relation. This turns out to be almost impossible, for two reasons. The shallow reason is the extreme ideologization of the debate over these values. The Business Software Alliance say:

Jobs in the Western European packaged software sector contribute six times as much to GDP as those in consumer goods.

and appear to have the figures to back this up. But I think few free software developers would trust anything the BSA says, and most other sources are likely to be similarly tainted on one side or the other of the divide.

The deeper reason links with the question of whether software can in the long run be successfully developed in a capitalist framework, and that is that money is simply not an effective way of measuring the value of software goods. The value of software is not something with an objective existence (in marginalist economics this is reflected in the impossibility of an optimal price for public goods). Since this runs counter to most common assumptions, it's worth taking it a bit further.

Most goods have a price and a value. The price is (relatively) simple: it's what shows on the sticker on the box. The value of a good to society might be estimated in different ways; one way is to estimate the amount of time (or work) needed to produce it. Assuming that the market is actually efficient at resource allocation in terms of society's needs, then the time spent on producing one good rather than another will actually be a reasonable estimate of the value of that good to society, and the price will be an approximation to the value. The assumption is obviously broken, but will do as a first approximation.

Packaged software clearly has a price, though the fact that this is a price not for the software itself, but for the license to use it is already a hint that there is something strange here. But what is it's value? The value of other goods sold on the market derives from the work taken to produce them socially, as part of the market system. The work is needed to reproduce the good, normally starting from an initial design. The design itself is a prerequisite without which the good can't be made, not part of its value. This is obvious with traditional goods; to take a random example, the great 19th century engineer Isambard Kingdom Brunel designed some of the worlds largest iron ships; without his skills they might well never have existed, but their value and their price were based on the number of workmen it took to build them, the amount of iron which went into their construction, and so on. With packaged software, the costs of production are tiny, not because of some magical property of information in digital form, but because the means of copying digital information are mass-produced on such a scale they have become very cheap. The same applies to analogue information once cassette recorders exist.

Packaged software therefore has a tiny value, not reflected in its licenseing price. That it has a license price above its value is made possible purely by the legal monopoly given by copyright law. The level of the price is also not arbitrary. Before a job could be carried out by software, if it was done at all it was done by people. Before word processors there were typing pools. The invention of the word processor made it possible to disperse the work done by the typing pools among all the staff of a company, including those without real typing skills, thus reducing company expenses by the wages of the typing pool. If the word processor licenses cost more than the typing pool wages, clearly the typing pool would keep their jobs. This is the first upper limit on the price. Secondly, once a type of software has become generalized, as is the case with the word processor, it is always possible to hire programmers to create your own version of a program. This is the second upper limit on the license price of software - the cost of hiring programmers to produce an equivalent piece of software. Of course, these upper limits are rarely approached: however monopolistic the industry, some competition is still possible, though it is not in the interest of competitors to push software licensing prices all the way down to the real value of the software.

Left at this there would seem to be a situation where a good is being sold consistently above its value, creating money from nothing. But the money does not come from nowhere - it is an overhead, subtracted from the profits of the companies that buy the licenses. Profits are being shifted from one sector to another.

OECD figures predict approximately $221.9 billion of packaged software license to be sold this year in the US, or 2.2 percent of the nation's $9.965 trillion gross domestic product. Microsoft has about $28 billion turnover, so it is still far from a true monopoly. Nevertheless from that $28 billion it expects a net income of $7.35bn. This is profit being sucked from other areas of the economy, partly within the US.

Once we look outside the US and other packed software producing countries, the situation changes. Money is no longer being shifted between sectors of the economy, but being removed from the economy altogether. From a company tax software licenses become a tax on the whole country.

So can this situation be changed by replacing packaged software with free software, product with service? In part this trend is already happening. Microsoft's potential competitors have already convincingly failed to compete with their own fractured Unices, their own GUIs, their own browsers. Increasingly their alternative is to package free software, which tends to become the only software which truly supports standards, and license it below Microsoft's prices. This has always been a possible tactic for smaller companies competing against a near-monopoly: offer standards and openness in an attempt to displace the monopolist.

Success in this tactic would mean the replacement of packaged software with free software. But at this moment the fundamental problem would reappear: packaging free software inevitably involves developing free software, and employing programmers (whether to write free or non-free software) costs money. The money cannot be recouped by selling software at its value, or by any purely market mechanism. Unless software sales are purely a marketing overhead for hardware sales, the model is not viable. Packaged software is generic software, which is precisely not a unique service: there is no need to produce individualised word processors or spread sheet programs for each company that uses them.

In the long run, the only way generic software can be produced within capitalism is through monopoly, with all the harmful effects this has: poor quality software, high social costs, forced upgrades, etc. What is worse, the only way that the monopoly can be maintained is to make the laws on which it is based ever harsher.

And in the meantime, dependence on free software by the capitalist sector leads both to a partial integration of the old firms within the free software mode of production, and a potential significant loss of control. If contributors to the software include those outside the firm, the boundaries of the firm begin to become blurred. It is no longer possible for managers within the firm to make arbitrary decisions about the form software should take, if this would lead to major objections from those outside the firm.

The New Economy

The old economy is stuck in a dead end as far as software production goes - it cannot develop without the new, but even with the new it is pushed into a sequence of ever harsher laws to survive.

The new economy, on the other hand has been dependent on the old from its beginnings. The dependencies correspond to the very features which make it possible to describe it as a new mode of production: production not for wages, yet programmers need money to survive; ownership of means of production, yet these means are affordable only because mass-produced within the old system; communication mediated by technology, not money, yet the technology is owned by some of the worlds largest companies.

Free software producers generally own their own means of production, the computers, CD writers, modems, etc but in most cases they don't make their own (and even those who do use standardised parts) which implies that computers cost significantly less than the average programmers income. This in turn implies either (or both) that they are being produced in very large quantities, or by very badly-paid labour. This bears repeating: free software production depends on mass production, and perhaps low wages too. This is a problem that cannot be solved within free software itself; nor is it likely that the solution will be a technical one. This dependence also introduces a possible threat: Microsoft, Intel and others can seriously consider hardware solutions to the problem of free software, Palladium being the most visible.

On the other hand, the dependence on technology owned by others for communcation may be avoidable. The development of local, free wireless networking is one possibility; and the past existence of FIDO where the Internet was not yet available shows that other solutions are actually practical.

Of the three dependencies, the dependence on money is the major one.

The need for jobs

Free software producers cannot live on their own produce. Until the free software economy has spread to more areas, they need money to survive, and only a very small proportion currently make this money by producing free software. Grouped by source of income, the three main categories of free software producer are student, employee, and self-employed, with a smaller category of people supported by various other means: unemployment benefits, inherited money, etc. The FLOSS survey showed 65% of respondents to be employees, 14% self-employed, 17% students, and 4% not working. This is a very high proportion of the self-employed; as a proportion of the total employed respondents, it is nearly 19%. This compares with a general rate of self-employment in the US of 8.6% in 1997, and 12.6% in the UK in 1999. On the other hand, in proportion to the population as a whole the percentage of students is also high; what is not known from the FLOSS data is what percentage of those categorized as 'employees' are in the university or state sectors or in the private sector. The FLOSS data certainly includes information which would provide support or disconfirmation for other aspects discussed in this section, but that data has not yet been released.

The three primary groups within free software development, arranged by source of income, would therefore be the self-employed and workers in small businesses; university and other state employees and students; and employees of larger companies. All three categories are likely to include at least some who develop free software at work, as well as in their own time. Further, at least two of the categories are quite permeable: in general, it is well known that self-employment has a negative correlation with unemployment levels, and many software developers move between employed and self-employed status at different times, depending in part on the state of the economy.

These three primary groups provide a possible explanation for some of the ideological differences among free software developers. Historically the first (pre-FSF) free software came from the universities (Spice, TeX, Colossal Cavern, etc) in the 1970s; and the FSF itself has always been strongly linked with MIT. Ideologies linked with this group (and state employees in general) are likely to be linked to support for state funding, education, and an emphasis on social justice. Given the length of state-funded education in Europe, these ideas may be more prevalent among Europe developers than in the US (perhaps reflected in the FLOSS survey's massive preference for Debian as distribution of choice; a result which would have seemed far more surprising in a purely US survey). The left form of this set of ideas is not so far from the ideas of traditional left parties, or newer parties like the Brazilian Worker's Party.

Large-scale self-employment in software really only became possible in the late 1980s and 1990s. Ideologies linked with this group are likely to be linked to freedom from state interference (and taxes in particular); in the US this has taken the form of Libertarian Capitalism - in many ways ideas that go back to the Levellers themselves. The left form of this set of ideas is strongly linked with anarchism; the independent shoemakers and cobblers who formed much of the core of Proudhonism in the 19th century being the classic example.

Though these two groups have major differences in themselves, as free software developers they have as much in common as they have separating them. The idea of 'freedom' as the overriding principle behind free software is a unifying factor, though in practice it may well be interpreted in different ways by the two groups.

The newest element comes from those involved in free software from within traditional companies, or employees of larger firms. This group are generally less likely to be ideologically motivated, and to emphasize the importance of quality of software rather than freedom.

Actual free software projects usually cut across the boundaries of each of these groups; as programmers, the economic situation of each participant may be irrelevant. Effectively, free software is based on an alliance of the three groups held together by the act of programming itself. And this is the core of the real washing-machine question: can such an alliance not only hold together but also expand to include other groups, groups who may not have any interest in programming itself?

Free Software and the Business Cycle

The most characteristic feature of capitalism is the business cycle. A new mode of production not dependent on the old system could be expected to be independent of this cycle. After all, an unemployed free software developer has more time available than an employed one.

But the dependence on work in the old economy does have some direct effects on the production of free software. Although the FLOSS survey found that 'Unemployment does not play a role within the group of OS/FS developers', the statistics for projects entered in Freshmeat show a slightly different story.

The last two years have seen a slump which has hit the software sector as badly as any; any plot of commercial software production over the period could be expected to show a sharp downturn. However, the figures for new projects added to Freshmeat under free licenses for that period actually show the number of projects entered per month as relatively static at around 500 per month.

(Note that figures before 2000 have been badly affected by the transfer of data from the old Freshmeat system, and are not reliable)

It is only when we look at projects which have since been updated - that is projects which have some life - that the effect becomes more obvious, with a slow decline from a peak in early 2000:

These figures are suggestive only, taking no account of the size of projects: Freshmeat counts Linux as one project, along with the smallest utility program. All the same, it does appear likely that free software still depends to some extent on the commercial cycle, mediated through the rate of unemployment.


In spite of these dependencies, free software production is already beginning to show its independence.

The key questions are whether the penetration of free software into the old economy is deep enough to start to slow the introduction of laws which can force it underground; whether free software can find allies outside itself to begin to reduce the dependencies listed above, and whether it can begin to gain new ground by moving into new areas, to gradually move closer to the washing-machine makers.

The first of these questions is the most urgent; but with the addition of IBM, HP, Sun and others to those who would be badly affected by harsher IP laws, they are no longer as inevitable as they seemed a few years ago.

But the second and third are the areas where there has been most movement over the last year. This comes in particular from those countries outside the charmed circle of official packaged software production. Here the rising levels of licensing fees (which, as pointed out above, are of no productive use whatsoever in the country) combined with pressure from Microsoft to enforce copyright laws (leading in some cases (Brazil, for example) to dawn raids, arrests, and mass confiscation of equipment) have begun to generate wider sympathy with free software.

Starting in Mexico, bills favouring the use of free software by governments have spread through Latin America: Mexico, Brazil, Peru, Argentina, Ecuador, Colombia. From Latin American countries they have spread back to Europe: Spain, Italy, Portugal in particular. And now such bills are appearing everywhere: the Ukraine, Malaysia, India, even California itself. Free software practices are visibly being carried over into the creation of the bills themselves: from the first relatively badly worded and poorly argued attempts, to a cumulative increased depth of understanding of their purpose and the best way to support that purpose, organised by people in contact across national borders who can take the lessons of each succeeding country with them.

Initially ferociously attacked by Microsoft, Microsoft now appears to be losing the initiative as the epidemic becomes too wide to contain. In part this is simply because Microsoft have found it more effective to spread FUD about these bills through local employers software organizations (real, or sometimes, created for the purpose) than to make it clear that they are the principal opponents of the bills. The most successful have not been at national level, but at regional level, partly because they sometimes manage to slip by under Microsoft's radar: Rio Grande in Brazil, Bologna in Italy, Estremadura in Spain; partly because they can count on a real basis of local free software writers and organizers, without which there is a risk that, as in Mexico, these will turn into largely paper exercises.

Each of these successes has begun the creation of new specialised software, which in turn will make it easier for others to follow in their footsteps. The arguments in favour of free software in the state are very easy to make, but its importance is not only it terms of logic and fairness (the ability of all citizens to access data held by the state, freedom from fear of back-doors, the public ability to assess electoral software for possible corrupt practices) but perhaps even more in increasing the pool of free software developers, providing additional employment (itself the principal argument in favour of the bills in some of the countries), allowing poorer countries to compete in software at a world level, reducing the ability of governments to casually pass laws restricting programmers freedom, and bringing new groups of people (such as state-employed teachers) to use and depend on free software and the working methods associated with it.

If these bills begin to pass as laws at national level, this may imply a shift in the centre of gravity of free software from the most developed countries to the less. It also implies a general shift in the political balance around free software: these bills are naturally proposed by left parties within each country, generally with views far to the left of the majority of American free software developers. This opens up the possibility of more links outside software production itself; if the Brazilian PT supports both free software in the state and production of farm produce in farmers co-operatives, what more natural than that the methods of one begin to rub off on the other? Of course, it also opens up the possibility of a split between free software developers and users. One reassuring sign here is that the most left-wing of all the parties to propose a free software bill (the Portuguese Left Block) when asked why they had proposed the bill, answered that it was due to pressure from free software users within the party.

The next new area is actually an old area: the universities. Once the defenders of the right to share research results, programs, and other works, through the years in many countries a change to a dependence on funding from the very companies that wish to impose IP laws, combined with government pressure to self-fund through patenting and sale of all their products has created a culture of closure rather than openness. The positions have reversed: once one of the inspirations for free software, free software is now an inspiration for many in universities who wish to regain their old position. Most visible in the original bastion of free software, MIT, where the OpenCourseWare program has begun in defiance of the trend to commercialising classes and lecture notes, the grounds well in favour of free software principles is growing throughout higher education. One of the most startling results is that the defenders of public access to the results of gene research do so not in the traditional terms of academic freedom, but using the rhetoric of free software:

The worm project [study of the growth of the nematode worm] worked so well because the community held an ethos of sharing - just as the public genome projects have - from the beginning. We gave all our results to others as soon as we had them. From sharing, discovery is accelerated in the community.

Sir John Sulston, Nobel Prize winner

Not surprisingly, Microsoft has reacted to these tendencies with the demand that universities use only BSD-type licenses for their free software. More surprisingly, the FLOSS survey (Part 2, final para) concludes with the same demand. In this case Bruce Perens' Sincere Choice platform which gives justifications for using either GPL or BSD style licenses in universities and concludes that this should be at the choice of those involved, seems the only fair one. But the very fact that Microsoft is now concerned about this is a sign of the growing hegemony of free software over wider areas of society.


Free software is a social as much as a technological development, and it will continue to spread by social means. A search for technological magic solutions to the 'washing machine problem' is a red herring; free software is already spreading beyond its original social bounds. The spread of bills on the use of free software in developing countries is one sign of this; another is the change from a rhetoric of academic freedom to one of the right to share and common development in the universities. The solutions to the washing machine problem will appear with the spread of free software principles and their implementation in still new areas.

Copyright 2002 Graham Seaman

Copying and distribution of this entire article are permitted without royalty in any medium provided this notice is preserved.

A. Bibliography