Build it!... but what if they don't come? J. Scott Marcus, Dieter Elixmann 1
The research question It is fairly clear that demand for broadband can drive supply. How does supply affect demand? Will abundant supply stimulate creation of new bandwidth-hungry applications? Alternative hypothesis: - too little available bandwidth can reduce consumption, but - too much bandwidth availability has only limited ability to promote bandwidth consumption. 2
Build it!... but what if they don't come? What is known about bandwidth demand? Three case studies: - The dot com bust of 2000 - The transition from dial-up to broadband - The deployment of fibre-based NGA in Japan and South Korea Observations 3
Internet traffic growth Important to distinguish between: - Growth in the number of users - Growth in traffic per user 4
PB/month Increase in Internet Traffic Global Internet Traffic Growth 70% 25.000,0 64,6% 60% 61,1% 54,4% 20.396,2 20.000,0 50% 47,9% 40% 14.685,6 38,9% 15.000,0 30% 9.928,1 10.000,0 20% 6.430,9 3.991,9 5.000,0 10% 2.425,6 0% 2005 2006 2007 2008 2009 2010 0,0 Total Growth Source: OECD Communications Outlook 2011 data, based on Cisco VNI, our calculations 5
millions Per 100 inhabitants Increase in Internet Users 2.500 2.000 Global numbers of individuals using the Internet, total and per 100 inhabitants, 2001-2011 Individuals using the Internet (in millions) Individuals using the Internet per 100 inhabitants 100 90 80 70 1.500 60 50 1.000 40 30 500 20 10 0 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 0 Source: ITU World Telecommunication /ICT Indicators database 6
Traffic per user (busy hour) Source: Cisco VNI 2011 data, our calculations 7
The dot com bust The most obvious counter-example to the build it and they will come maxim is visible in the US dot com bubble/bust. During the 1990s, the boom saw rapid deployment - of long haul fibre infrastructure, - of metro fibre rings in larger cities, and - the deployment of substantial transatlantic capacity. Beginning in March 2000: - dramatic fall of technology stocks, - disappearance of many Internet firms, - Collapse or retrenchment of many prominent telecommunications network operators including MCI WorldCom; Global Crossing; Qwest; and Genuity. 8
The dot com bust Resulted in a massive over-supply of long-haul fibre in the United States. For several years after the dot com bust, long distance transmission capacity could be acquired for a small fraction of what it had cost to build the capacity. Suggests that there must be an upper limit to the ability of supply to stimulate demand. 9
Shift from dial-up to broadband Traffic statistics prior to 2005 are uneven. We reviewed data on shipments of service provider routers, and of long-haul DWDM, as potential proxies for traffic growth. Business cycle effects (the economic crisis circa 2009) loom large. A sophisticated analysis would be needed to see if any effect is visible in the data. 10
Shift from dial-up to broadband 6.000 Worldwide high end service provider router capacity in Mbps 3.500 Long haul DWDM capacity shipped (Mbps) 3.000 5.000 2.500 4.000 2.000 3.000 1.500 1.000 2.000 500 1.000 0 2004 2005 2006 2007 2008 2009 2010 2011 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 Long haul DWDM capacity shipped (Mbps) Dell Oro data, our calculations 11
Fibre-based NGA in Japan and South Korea If an effect is visible anywhere, it ought to be visible in Japan and South Korea. 35.000 30.000 25.000 20.000 15.000 10.000 FTTH DSL Cable 5.000 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Source: Tatsuo Takita, Infocom 12
GB/month per subscriber Japan and South Korea vs Europe 120,0 Traffic per subscriber Internet Traffic per Fixed Subscriber 100,0 80,0 60,0 2011 2012 40,0 20,0 - Japan South Korea Germany United Kingdom Italy France Data sources: Cisco VNI 2012 online database, OECD, ITU, own calculations 13
Japan vs Europe Source: Rupert Wood, Analysys Mason 14
Concluding observations Results are not definitive, but... None of the coarse indicators that we looked at substantiates the existence of a strong build it and they will come effect. The Japan data support the alternative view. The South Korea might support the existence of a build it and they will come effect, but weaker than many have assumed. 15
The good news Methodologically, this suggests that models of broadband demand are not likely to go far astray if they treat broadband demand as being for the most part exogenous (i.e. largely independent of supply). This is a useful simplification. 16
The broader implications? If ultra-fast broadband really generates societal benefits, then why are those benefits not more visible in the way in which consumers use their broadband connections? Are the NGA networks that have been deployed in Japan to date (presumably at great cost) largely empty in terms of traffic carried? Consumer incremental WTP for ultra-fast broadband appears to be no more than 5/month. Is it possible that consumers are doing a better job than policymakers when it comes to assessing the incremental benefits (today!) of ultra-fast broadband? 17