Tuesday, August 24, 2010

Broadband Internet and economic growth: some thoughts

There's a lot of head-scratching over net neutrality and innovation - and the effects of centralized versus localized innovation on economic growth. On the local side, there are Seely Brown, Benkler, van Schewick, Lessig, and economists Economides, Evans and Brynjolfsson. On the central side, there are any number of managerial experts, and its certainly the case that previous industrial revolutions depended to a large extent on the management of capital and resources, notably in warfare and railroads.
Andrew Odlyzko's latest work has set me thinking, along with Christian Sandvig's reminder of the brutal corporate strategy of the railroad and oil pipeline bosses of the Gilded Age (which I back in the bubble of 2000 decided had been replaced by a new George Gilder-ed Age).
John Seely Brown has long talked of the role of information processing in terms of enabling technologies (or general purpose technologies), and of the second forty year period as more important than the first in any enabling technology's development. This is not new - it is known as Solow's Productivity Paradox - and reflects the socio-economic lag before the full benefits of new technologies can become generally deployed in any market. Hence, the steam engine was invented in 1774, the mobile version - steam trains - in 1805, the iron-clad ocean-going ship in 1843 (and the enormously advanced SS Great Eastern in 1858, largest ship for over 40 years), but their effect was far greater over the third and fourth human generations after their invention (note that Brunel's great ships became respectively a gold rush passenger carrier to Australia and a telegraph cable carrier, thus enabling enablers). The same process applies to the motor car, the microprocessor, electrification and so on.
So with the Internet. It was 'invented' (depending on your taste) in about 1968, so it is now entering its third generation with widespread consumer adoption of broadband.
Its worth adding that these enabling technologies can produce bubbles in investment because their development is so lumpy before it becomes entirely ubiquitous - and Odlyzko's thesis is in part that this bubble speculation is in the long run beneficial for development (if not for investors) as it spurs that ubiquity. Hence the railroads have seen multiple bubbles in investment, which eventually resulted in oversupply of track in the UK, but phenomenal growth in economies which copied that development (as well as in the UK itself). Argentina, Paraguay, Siberia, Utah, the Punjab, Kenya, all benefited enormously from the railroad up to a hundred years after its invention. In consequence, beef, wheat, rice, cotton markets all made huge strides, as did real estate, tourism, postal services, local road transport and so on.
What does this mean for net neutrality? Well, first is that real damage to wider innovation, should it occur, may not matter much until 2050 or so, at which point it will have utterly shattered the lead of .
Second, the core network usages may simply resist innovation - railways have more or less used standard gauge ever since 1850, despite the attractions of broad gauge and Brunel's other innovations. An unmanaged pipe may be best - just as Brunel's 130km/h trains of the 1870s barely needed speeding up for 100 years (and are still timetabled speeds today!). But Japan's bullet trains from 1964, and France's TGV from 1981, are only now being replicated in Germany, China and so on, and the UK is unsurprisingly at least half a century behind with its fiendishly useless railway regulation almost 20 lamentable years old, and no investment or longterm planning.
Third, the network itself may not need to continue to exponentially increase its speed: perhaps Ethernet is fast enough, or at least can be trivially upgraded for consumers. Of course you have to reach Ethernet speeds first by fibre to more or less everyone in the country. So once you're up to speed and have several alternative distribution networks (e.g. rail, canal, road, bicycle/horse now outmoded), you don't need to invest in some maglev type of super-fast service. Take an example: satellites are brilliant at delivering high definition television, mobile spectrum is great for telephony. The fixed Internet doesn't need to do everything.
Fourth, you could fiddle around the edges ('managed services' anyone?) just using the spare physical capacity in the ducts, as railways developed telegraph and telecoms services next to the rail lines. Its what keeps trunk telecoms working in most countries even now - MCI in the US used the railways' rights of way, most UK companies used the same - after all, they had 17,000km of fibre ready for competitors to BT.
But at the centre is the unrebuttable claim that the innovation caused by a largely distributed architecture is the best for innovation - it cannot be rebutted because we don't have enough convincing evidence to prove the contrary, unless one simply maintains that telecoms companies are hopeless innovators and thus should be kept away from fiddling with the system.
However enabling the Internet will become - and it promises a great deal - it will achieve a great deal more in the next fifty years. Whether it will be allowed to do so is what net neutrality is all about, whether you frame the issue in technical, social, economic, political, philosophical or any other terms. A bubble of investment that allowed developed countries to simply run fibre to their customers would solve most of the connectivity part of the problem. How what gold rush fable can we invent to make it happen? How about Free's business model, an all-you-can-eat buffet at 30euros with no sales: the world's greatest ISP?

2 comments:

Richard Bennett said...

Contrary to popular belief, there is nothing about communication networks that promotes decentralization and local decision making. To the contrary, high-capacity networks simply extend the control that a centralized system can affect, and make remote decision systems more effective. Monitoring, sensing, and information gathering take place at the edges of the Internet, but databases are in the centralized computing complexes euphemistically called "The Cloud."

Lessig and his acolytes sell a great deal of what amounts to young adult science fiction, but I fail to see how anyone can regard them as serious thinkers.

You make dig at managed services, which I think probably reflects a lack of understanding about how they're used today and why they're essential. One common use of managed services is to enable high-resolution video conferencing, both within firms and between firms. They buy managed services from telcos and from independent network operators. This is necessary because an end-to-end, best-efforts Internet can't eliminate jitter, and therefore can's support this mode of interaction.

Finally, if people are still using the Internet in 50 years, that will be a total failure of innovation and a tragedy of epic proportions. I so hope you're wrong.

chris said...

I was actually making claims for the transformative effects of 'managed services' but note that my definition is of services provided outside the Internet pipe. Whether provided on the side of the Internet pipe or via a different method, such services will be an important part of the mix.
I think you misunderstood 'managed services' as Quality of Service, a quite different discussion which I do not enter into in this blog post.
'Tinkering' around the edge of the network is what made the Victorian innovations great - hence my SS Great Britain and telegraph references. Use the common carrier to innovate on the right of way.
As people still use the telephone, Morse code and radio, so they will still use the Internet. Ubiquity and a common standard have a habit of entrenching themselves. Whether its the dominant protocol is more open to debate.