VoxEU Column Labour Markets

The geographic concentration of services today mirrors that of manufacturing a century ago

Recent research suggests that improvements in information and communication technology are affecting the spatial growth of services today in the same way electricity promoted the concentration of manufacturing at the turn of the 20th century.

In the last two decades, rapid improvements in Information and Communications Technology – ICT – has wrought enormous change on the world. As with other so-called general purpose technologies, its productivity impact was initially not much felt. As Solow quipped in 1987, computers were everywhere “except in the productivity statistics”. But with the passing of time, the potential of ICT unfolded. The productivity figures started to pick up. One concrete example of the benefits from ICT is the recent growth in outsourcing and offshoring of different tasks. Now that “everything you can send down a wire is up for grabs”, in the words of Nandan Nilekani, the CEO of Infosys Technologies, it would appear that distances no longer matter, and that any task can be performed anywhere.

However, evidence from the US suggests that announcing the “death of distance” may be premature. If distances were to cease to play a role, then economic activity should become more equally spread across space, as firms and workers move from areas with high land rents to areas with low land rents. In line with this prediction, manufacturing employment inside the US has de-concentrated in recent decades; US counties with small manufacturing employment have experienced faster than average manufacturing job growth. The pattern for services, however, looks very different.

Except for very small and very large counties, the last 20-30 years have witnessed increasing spatial concentration in services. Focusing on intermediate-sized US counties, the larger ones, in terms of service employment, have seen faster growth in service jobs. The picture below shows how, between 1970 and 2000, manufacturing became increasingly de-concentrated, whereas the different service sectors, such as retail, exhibited an S-shape growth pattern, with growing concentration for intermediate-sized counties.



This starkly opposing pattern in manufacturing and services does not sit easily with the idea that rapid advances in ICT are edging distance towards its death bed. Services, after all, would seem to be more intensive in things that can be sent down a fiber optic cable. And you cannot send cement through a wire. To put it differently, the service sector has been going through a technological revolution; manufacturing has not. When measuring the age of an industry as the time elapsed since it was last impacted by a general purpose technological innovation, the differential impact of ICT allows us to label the service sector as young and the manufacturing sector as old.

It makes sense to think that the diffusion of general purpose technologies is not indifferent to the spatial concentration of economic activity. For a new general purpose technology to reach its full potential, much experimentation and many smaller subsequent innovations are needed. Locations at the technology frontier benefit from agglomeration economies. Firms exploring the multiple uses of the new general purpose technology want to be geographically close to other firms doing the same. Once these incremental innovations are developed in spatial clusters à la Silicon Valley, they then diffuse to other regions of the economy.

This logic helps explain how agglomeration economies and technological diffusion produce the S-shaped growth pattern for young industries, such as services. Start by considering the very small counties in terms of service employment. As innovations diffuse, some counties previously producing in other sectors start specialising in services. This implies that counties with initially low-service employment can experience fast growth in that sector. However, once we get to intermediate-sized counties, agglomeration economies become important. The growth pattern reverses, and concentration takes over. Once counties become really large, though, congestion kicks in, and technology diffusion no longer leads to growth. Mature industries, in contrast, have long passed the experimentation and diffusion stage, and have standardised technologies. They no longer benefit from agglomeration economies, and prefer moving to locations with lower land rents. Employment growth therefore exhibits negative scale dependence in the largest countries.


If this logic linking spatial concentration and the age of an industry is right, then manufacturing at the beginning of the 20th century should exhibit a similar pattern to services in recent decades. The last important general purpose technology to affect manufacturing was the introduction of electricity. Its diffusion has been traced back to the period 1900-1920. The advent of electricity had similar effects as ICT on, say, the stock market, IPOs, industry shakeouts, and patents (Jovanovic and Rousseau, 2005). The same is true for spatial concentration patterns. The picture above shows how manufacturing over the period 1900-1920 looks just like services in recent decades. The overlap of the different S-shaped curves is nearly perfect.

This finding is not limited to the United States. In recent decades European regions have also witnessed S-shaped growth in services and increasing dispersion in manufacturing.

Neither is our finding limited to the distinction of manufacturing versus services. After all, within the different service sector, certain sub-sectors are ICT-intensive (say, banking), whereas others are not (say, haircuts). Of all the service sectors, the data tell us that ICT is having the most impact on legal services, and least impact on auto repairs. Not surprisingly, we find S-shaped growth in legal services, and de-concentration in auto repairs. In other words, auto repairs behave as a mature sector, not affected by ICT, just like manufacturing.

Although ICT has been hailed as bringing about a new economy, what we are witnessing may not be that new after all. The spatial growth of the industries it affects looks much like that of the industries affected by electricity at the beginning of the 20th century. If history is a guide, then as ICT further diffuses and matures, agglomeration economies and diffusion will weaken, and services will start to disperse in space.


Jovanovic, B. and Rousseau, P.L., 2005. General Purpose Technologies, NBER Working Paper #11093.


This commentary is based on “Spatial Growth and Industry Age” by the same authors available as CEPR Discussion Paper No. 6421.

420 Reads