End of the office?

In this piece in the New York Times Edward L. Glaeser asks,

Will electronic connections make cities obsolete? In the giddy early days of e-mail and the Internet, some prophets proclaimed that humans would no longer bother with the inconveniences of density and would instead retreat, in Alvin Toffler’s phrase, to “electronic cottages.”

This move to the "electronic cottages" raises a related question, Would the "electronic cottage" spell the end of the office or the factory? The issue has to do with the determination of whether or not work occurs in centralised factories or in separate households or some combination of the these. But to be fair this is nothing new, this has been an issue since at least the industrial revolution.

In his discussion of the development of the factory system during the industrial revolution, Mokyr (2001, 2002: chapter 4) puts forward the argument that the location of production was dependant, in part, on the trade-off between “the relative costs and benefits of moving people as opposed to moving information.” Mokyr (2002: 120). That is, he develops a line of reasoning that suggests that one factor encouraging the organising of workers under a single roof, rather than in separate households, was the division of knowledge. As long as there was little division of knowledge, so that the knowledge needed to carry out production could be summarised in a few basic rules, the household could know all that was needed to act as the “unit of production”. The cost of transferring information between workers was low since there was little of it needed and the workers were contained within the household. Moving people between households or to factories, however, was slow and costly. But as technology developed, the competence required for production moved beyond the capability of a single household. As the knowledge needed to produce things increased, firms faced two related problems. First, a firm needed to be able to incorporate existing knowledge into their production system and second, they had to be able to generate new knowledge to keep or establish a competitive advantage. Inevitably, specialisation and the division of labour became finer. The way to deal with the increased level of knowledge demanded for production was to divide up the production process into smaller manageable tasks. Workers had knowledge about ever smaller pieces of the production puzzle. A result of this more extensive division of labour, which could in some circumstances reinforce the movement towards a single location factory, was noted by Babbage (1835). Babbage observed that the greater the division of labour the less time required for learning any requisite skills. This results in a lessening of the period during which a new entrant to the workforce would be relatively unproductive and unremunerative. Because less knowledge and training was required to learn to undertake a single operation, as opposed to that required to undertake many different operations, a new employee would more quickly reach a situation where he generates a profit for his employer.

Mokyr points out that the importance of the division of knowledge to the firm was first recognised, albeit in a non-historical setting, by Demsetz (1988) and formalised by Becker and Murphy (1992). What these works suggested was a new interpretation of the role of the firm. Given that there are limitations to what a worker can know, the competence that a firm has to possess to produce must be divided into manageable portions and allocated among the workers. The actions of the different groups of workers are then coordinated by the firm’s management. Thus, workers who produce on the basis of knowledge they themselves do not possess, have their activities directed by someone who does possess (at least more of) the necessary knowledge. Therefore, the coordination needed due to asymmetries in information among workers provides a rationale for management. In this way, direction is a substitute for education, that is, a substitute for the transfer of the knowledge itself. Specialisation in knowledge can, therefore, both exacerbate existing information asymmetries and create new ones. Any information asymmetry gives rise to an organisation problem for the firm: How can agents who possess knowledge be encouraged to reveal their knowledge fully and truthfully to other workers or management? Mokyr argues that:

Putting all workers under one roof ensured repeated interaction and personal contact provides maximal bandwidth to maximize the chances that the information will be transmitted fully and reliably. Inside a plant agents knew and could trust each other, and this familiarity turned out to be an efficient way of sharing knowledge.” (Mokyr 2002: 141).

From this, it can be seen that as long as the minimum competence needed by a firm is small, the plant size can also be small and can, therefore, coincide with the household. When the competence needed grows, the unit of production has to change or an efficient network for knowledge distribution has to develop. At a time when the main technique for the distribution of knowledge was direct contact, as at the time of the industrial revolution, such networks in the form of professional associations of mechanics, machinists, engineers etc did develop. But the firm was also an answer to the problem of knowledge distribution. Costs of accessing knowledge were minimised in a single plant where workers could communicate face to face. Factories acted as repositories of technical knowledge and allowed workers to access this information at relatively low cost. Thus factories allowed knowledge to pass in two directions: across space so that other workers could carry out a given task and through time so that knowledge passed from one generation of workers to the next.

McDermott (2001: 48) explains that the transition to factory life altered the incentives for both owners and workers with regard to acquiring knowledge and providing training. Workers now had the opportunity to acquire highly specialised knowhow about both their particular firm’s production process and about more general, transferable, skills. Owners now had an incentive to educate their workforce since firms unlike households can go out of business if they do not keep up. This gave owners the needed impetus to train their employees. Galor and Moav (2006) also argue that support from capitalists for the education of workers was due to the increasing importance of human capital in sustaining their profits. Physical capital accumulation in the process of industrialisation gradually intensified the importance of skilled labour in the production process and thereby generated an incentive for investment in human capital. Due to the complementarity between physical and human capital in production, the capitalists were among the prime beneficiaries of the accumulation of human capital by workers. By putting workers under one roof, it was easier for owners to compare workers’ productivity and select those who were most suitable for greater training and advancement. This process helped differentiate the firm from the household.

Differences in knowledge between principals and agents can affect the desire to move to a centralised factory for additional reasons. Lamoreaux, Raff and Temin (2003) note that the putting-out system in US cotton-spinning came under pressure, in part, because of principal-agent problems between the manufacturers and home based weavers. Lamoreaux, Raff and Temin (2003: 412-3) write:

However, the enormous coordination problems that this system entailed (for example,
unsupervised weavers working in their homes turning out fabrics of vastly varying
qualities) spurred manufacturers to reconcentrate production in factories as soon as
technological innovation in the form of the power loom enabled them to expand capacity sufficiently.

Kim (2001) suggests another way in which information can affect firm organisation and business location. Kim makes the point that specialisation adds to transaction costs via a loss in information. With specialisation, firms know their costs but are uncertain as to the demand for their products while consumers know their demands but not the supply conditions of production. Kim notes that before the late nineteenth century, most goods were produced using craft technology and were often produced in homes or at best small shops. Such firms tended to operate in a local or regional market, produced a single line of output, and were owned and managed by a single individual or a partnership. Given the relative simplicity of production, consumers could identify the quality level of products either via physical inspection or through the reputation of the producer or seller. As the production process became more sophisticated and production took place away from the consumers’ location, information became more and more asymmetric. Producers knew the quality of their goods but consumers were less well informed. This gave raise to a potential adverse selection, or “lemons” problem, where bad products drive out good ones.

Kim’s argument is that the modern multiunit firm is a solution to the asymmetric information problem. So production moved away from single-plant, single-region, single-product firms towards multi-plant, multi-region, multi-product enterprises to counter potential lemons problems. The advantage that the multiunit firm had was that as repeat sales were of greater value to the multiunit firm, they were better able to make large firm-specific, sunk cost investments in advertising and branding to credibly signal to buyers that the costs of reneging on quality were high. Kim (2001: 311) writes:

Multiunit firms were able to solve the asymmetric information problem through the use of advertising and the development of brand names. In the presence of uncertain quality and the absence of a credible third party enforcer, the main private-contract enforcement mechanism relies on the value of repeat sales to a firm. One solution to signaling a firm’s value of repeated sales is to invest in firm-specific and non-salvageable assets such as advertising and developing brand names. Since the value of repeat sales is limited for most single-unit firms, these firms have little incentive to advertise and develop brand names. On the other hand, for multiunit firms, the value of repeat purchase is potentially much greater. Thus, the economies of marketing for multiunit firms come not only from their ability to spread their costs over many plants or stores, but also from the fact that the cost of reneging on their product quality is
significantly higher.

Audia, Sorenson and Hage (2001: 79-82) see information affecting the location decision in other ways. For them, there is a tradeoff in the organisation of production between geographic dispersion and organisational learning. First, a multi-plant firm has advantages in the creation of new knowledge with regard to the optimal method of production. With multiple sites, even random variation across those sites can offer the opportunity to gather comparative information on the best production methods. But more strategically, multi-unit firms can undertake parallel experimentation with different plants undertaking different experiments at the same time. This offers at least two advantages: 1) It allows learning to occur at a faster pace. As experimentation takes place in chronological time, being able to undertake multiple-experiments simultaneously reduces the time needed for the firm to investigate the benefits of changes in operating procedures; 2) With the greater number of observations, the internal validity of any conclusions is greater. So geographically dispersed firms benefit more from knowledge creation via experimentation than geographically concentrated firms. Second, geographic dispersion of plants will affect the efficiency of knowledge transfer. Here the single-plant firm has an advantage − for reasons similar to those put forward by Mokyr for the development of the factory during the industrial revolution. In particular the transfer of tacit knowledge is difficult even with face-to-face contact but without it such a transfer may be nearly impossible. A second issue with geographic dispersion relates to the usefulness of the knowledge being transferred. As the similarity of the plants declines, the usefulness of information learned at one location to other locations is reduced. In short, geographically dispersed firms benefit less from firm knowledge than geographically concentrated firms.

As pointed out by Mokyr (2002: 141), the Demsetz/Becker-Murphy framework also suggests that when knowledge can be shared and believed among agents without the need for personal contact then firms may survive, but the large plants we know today may become less necessary. This point is becoming more important as the use of ICTs is expanding.

The development of ICTs has meant that the costs of moving people as opposed to moving information have risen sharply. The costs involved in sending and receiving information have fallen thanks to technologies such as email and the Internet along with falls in the costs of long-distance phone calls and the expanding use of cellular networks. The costs of people moving have not fallen so dramatically. Commuting to work via congested city and suburban streets, for example, is at least as difficult as it was two decades ago. The increasing interest in congestion pricing in many cities around the world suggests that traffic problems are not lessening. The increasing relative cost of moving people as apposed to information would suggest that the size of the “unit of production” should be moving away from the large factory, so dominant for the last two centuries, towards more home based production, as in the period before the industrial revolution. Mokyr (2002: 155) does however add a cautionary note. He argues that the movement away from work in the factory setting will at some point run into diminishing returns and what we will see is the locus of work remaining a mixture of work at home and work away from home.

McDermott (2001: 52-3) reinforces this cautionary note by raising four issues. First, McDermott argues that monitoring remotely would be problematic given that to be effective, it may violate the norms of privacy. Secondly, joint production in the home of market and household goods could diminish in the future. The increased market provision of “household goods” − day care, mobile dog grooming, internet shopping and home delivery etc − all allow workers to spend an increased amount of time away from home. Thirdly, the growth in personal services means that in the future many workers will have to serve people directly in ways that would be difficult to do remotely (personal trainers, customer service representatives, mechanics, craftsmen for example). Last, to quote McDermott (2001: 53): “As Mokyr notes, citing Gavin Wright, “In the limit we could devise an economy in which technology is designed by geniuses and operated by idiots”. In that case, home production for market may indeed take off. But I am uneasy with that conclusion. It seems to me that a large part of the population may, unfortunately, substitute information and computing power for their own brain development, but these workers will not be the kind that will be left alone to telecommute. These workers will require considerable oversight in something like a factory.”

The Demsetz/Becker-Murphy model indicates that when knowledge is an important factor of production, small firms have advantages. If Mokyr is right, then this downsizing of firms should lead to a movement back towards home production and away from large factory production. But even if this is so, it is not clear whether these home producers will be single-unit firms or units of a multi-unit company.

For Glaeser face-to-face interactions and electronic connections are at least in part complements and thus large cities have big advantages. The same can be said at the more micro level of the factory or office. Changes in technology will not spell the end for either the city or the factory or office within those cities.