Monday, November 17, 2014

Andrew Odlyzko Strikes Again

Last year I blogged about Andrew Odlyzko's perceptive analysis of the business of scholarly publishing. Now he's back with an invaluable, must-read analysis of the economics of the communication industry entitled Will smart pricing finally take off?. Below the fold, a taste of the paper and a validation of one of his earlier predictions from the Google Scholar team.

Among his observations are:
  • "by some measures the US spends almost 50% more in telecom services than it does for electricity."
  • Content is not king; "net of what they pay to content providers, US cable networks appear to be getting more revenue out of Internet access and voice services than out of carrying subscription video, and all on a far smaller slice of their transport capacity".
  • True streaming video, with its tight timing constraints, is not a significant part of the traffic. Video is a large part, "but it is almost exclusively transmitted as faster-than-real-time progressive downloads". Doing so allows for buffering to lift the timing constraints.
  • "The main function of data networks is to cater to human impatience. Thus "Overprovisioning is not a bug but a feature, as it is indispensable to provide low transmission latency". "Once you have overengineered your network, it becomes clearer that pricing by volume is not particularly appropriate, as it is the size and availability of the connection that creates most of the value."
  • "it seems safe to estimate worldwide telecom revenues for 2011 as being close to $2 trillion. About half the revenue ... comes from wireless."
  • "with practically all [wireline] costs coming from ... installing the wire to the end user, the marginal costs of carrying extra traffic are negligible. Hence charging according to the volume of traffic cannot easily be justified on the basis of costs.
  • "a modern telecom infrastructure for the US, with fiber to almost every premise, would not cost more than $450 billion, well under one year's annual revenue. But there is no sign of willingness to spend that kind of money ... Hence we can indeed conclude that modern telecom is less about high capital investment and far more a game of territorial control, strategic alliances, services and marketing, than of building a fixed infrastucture."
  • "Yet another puzzle is the claim that building out fiber networks to the home is impossibly expensive. Yet at the cost of $1,500 per household (in excess of the $1,200 estimate ... for the Google project in Kansas City, were it to reach every household), and at a cost of capital of 8% ..., this would cost only $10 per house per month. The problem is that managers and their shareholders expect much higher rates of return than 8% per year. One of the paradoxes is that the same observers who claim that pension funds cannot hope to earn 8% annually are also predicting continuation of much higher corporate profit rates."
Back in 2002, Odlyzko analyzed the usage of online content through time after its publication. Initially, the decay was rapid but after a while usage settled to a low constant level or increased. On this basis he predicted that there would be much wider citation of older articles.
Of the articles that were most frequently downloaded [from First Monday] in 1999, 6 of the top 10 were published in previous years! This supports the thesis that easy online access leads to much wider usage of older materials. [Section 9]
After an initial period, frequency of access does not vary with age of article, and stays pretty constant with time (after discounting for general growth in usage). [Section 10]
Now The Google Scholar team have followed their Rise of the Rest paper, which I blogged about here, with a validation of Odlyzko's prediction. Their new paper On the Shoulders of Giants: The Growing Impact of Older Articles takes another look at the effect that the dramatic changes as scholarly communications migrated to the Web have had on the behavior of authors. The two major changes have been:
  • The greater accessibility of the literature, caused by digitization of back content, born-digital journals and pre-print archives, and relevance ranking by search engines.
  • The great increase in the volume of publication, caused by the greatly reduced cost of on-line publication and the reduction of competition for space.
The paper shows that in most fields, the proportion of citations to articles more than 10 years old has increased significantly (28% to 36% overall) from 1990 to 2013. The same holds true for 15 and 20-year old articles. The rate of increase is accelerating. There are some outliers, Chemical and Materials Science and Engineering excluding Computer Science both show little change. Computer Science, on the other hand, shows a significant increase, but this is bi-modal, 5/18 of the CS subject categories show less than 30% increase whereas 11/18 show 50% or more.

No comments: