Will mobile internet `crowd out’ fixed line internet?


According to Phil Dobbie, the answer is clearly ‘no’. See here. And Phil’s case is strong, that fixed-line internet, with its higher speeds, dominates in terms of ‘big downloads’:

Clearly, we’re using our fixed-line connections for heavy lifting, and that situation isn’t going to change anytime soon, if ever. Downloads from the average fixed-line user have increased almost 50 percent over the last year, but just 11 percent for the average mobile user.

But ‘ever’ is a long time and I am less sanguine that fixed-line will dominate mobile for ‘ever’, at least in its current form.

My caution is based on three bits of information. First, we have had about 20 years experience with fixed versus mobile telephone services. In developed countries, early studies showed that fixed line and mobile phone services tended to be complements. But more recently they have behaved like substitutes with fixed line subscriptions falling as mobile phone services increase. Ingo Vogelsang has a survey here.

While internet services and telephone calls are different telecommunications services, and ‘speed’ is not as important for voice calls as it is for internet services, initial limitations on mobile telephone services, such as cost and reliability, have declined over time, so that they are now a competitive alternative to fixed line telephone systems. Technology may (and probably will) close the speed gap between mobile and fixed line internet services over time.

Second, technology and innovation will be driven more and more by the demands of developing countries over time. In Asia, South America  and Africa, there are large populations who are going to become richer over the next few decades. These populations have been ‘brought up’ on cheap mobile phones. In part, this is due to the lack of adequate fixed line infrastructure in many of these countries to serve as a base for fixed-line services such as ADSL. So there will be big profits to be made from mobile innovations that can serve these markets. While the mature markets of Europe and North America will still matter, I suspect the innovation focus will become more ‘mobile’ oriented.

Third, there is the simple difference in infrastructure cost and speed of roll out. And this drives (and is driven by) competition. As our current NBN experience shows, it can take a long time and a lot of money to roll out high speed fixed line networks. And often these networks are monopolies. In contrast, the speed of roll out of successive generations of faster mobile internet has been impressive. I will let others judge how much this is driven by the different technology and how much it is driven by competition. But from an economic perspective, competition matters for service quality over time and mobiles are the ‘competitive space’ for internet infrastructure.

This suggests that, in developed countries, a likely future will be a hybrid. Fixed lines will provide digital carriage services between mobile ‘base stations’. Of course, this is what happens already for mobile telephone networks. And it is what happens on universities (campus wireless rather than wired buildings). But this is a very different future to one where every household or office has a wall socket to provide internet services.

Of course, this all feeds into current NBN policy. An NBN policy is a lot like crystal ball gazing. Rolling out fibre to the home is ‘flexible’ in terms of fixed-line options. It can be upgraded over time and should provide high speed services for a long time. But it is also a ‘wall socket’ approach. It is a bet on the Phil Dobbie view of the world. If this view is wrong, and in 20 years fixed line internet services are viewed as dinosaurs compared to mobile services, then much of the NBN will be a GWE (great white elephant).

11 Responses to "Will mobile internet `crowd out’ fixed line internet?"
  1. Wireless download speeds are limited by the amount of available spectrum and the Nyquist theorem. We are currently pushing up against those limits, and no amount of clever technology will get us past them. On the other hand, the theoretical maximum for optic fibre is, for all practical purposes, infinite.

    In any case, the main driver of demand for bandwidth is high quality video, which can only be viewed properly on devices with large screens. These devices generally use fixed connections, either directly or via WiFi.

  2. If you don’t really understand the mathematics underlying all radio technologies, then certainly it looks like there’s no reason to go wired over wireless in the long run.

    In reality though there are hard physical limits to the rate of information that you can transfer using a given portion of the radio spectrum. Certainly there are technological improvements that can be made, like spatial division using beam-forming techniques or smaller, lower-power cells – but ultimately you can’t scale spectrum availability to meet the number of users forever, in the same way that you can easily lay more fibres down.

  3. Mattkwan and caf – Thanks for the engineering background. Do the limits on information transfer rates for wireless imply a hybrid solution where there is fixed line to a local point and then wireless? For example, small base stations outside houses that provide a local network? The expensive bit of fixed line is getting into the buildings (e.g. getting to the house from the local exchange for fixed line telephony).

  4. You can get very good wireless speeds using high frequency radio, but those frequencies don’t penetrate walls very well. That limits to what you can do with street-level pico-cells.

    In any case, running fibre optic cable from a street-level node to each house isn’t *that* expensive. No more so than the Foxtel cable rollout.

  5. Mattkwan and caf, excellent replies, I was about to say the same thing. Essentially it boils down to how much information can be sent in any given medium, the noise and the strength of the signal. Bottom line is that we all use the same bandwidth for wireless, and there is only so much of it available, the channel is noisy, and signal strength drops quickly. Fibre optics doesn’t have those problems.

    Stephen, I’ve been reading a great primer on Information theory, including the roles of Nyquist and Shannon, who proved this stuff mathematically about 50 years ago. It’s about 30 years old now, but most of it still applies. http://www.amazon.com/Introduction-Information-Theory-Symbols-Signals/dp/0486240614.

    Bottom line is that there is a hard limit to the amount of information the air can carry, and the market won’t beat it any time soon. Fibre will be better than wireless for the forseeable future.

    To me, pushing back on the NBN because ‘in the future that will change’ is like stopping building roads, because soon we will have flying cars.

  6. You might want to check this with the engineering department, my comp eng. degree is a while back, but from memory, the core of the issue is the Shannon-Hartley theorem. Shannon built up a mathematical model of a waveform carrying information in a channel in the presence of noise, and showed that it’s information carrying capacity is proportional to its bandwidth times the log of 1 + the signal to noise ratio.

    As a mathematical model, it is equally applicable to lightwaves in fibre, radiowaves in air and voltage in copper wire. It also has a mathematical proof. Anybody who can prove it wrong, well they will be very, very rich, but I wouldn’t hold my breath.

    From this model we can explore some features of a theoretical ‘ideal’ medium. Firstly it isn’t shared, because if multiple people are using the channel then they share bandwidth, secondly it has very low attenuation of signal over distance, thirdly it has very low levels of noise. Anything that satisfies these criteria is a near ‘perfect’ technology.

    So, with this in mind, wireless is shared, there is a lot of radio noise from hundreds of sources both human and other, and the signal attenuates proportionally to the square of the distance from the antenna. None of these factors is likely to change with technological advances, all we will be able to do is work harder to remove redundancy from our information, but that too is very much a case of diminishing returns.

    Copper isn’t shared, but suffers from crosstalk noise as well as radio interference, as the copper wire acts as an antenna. The signal attenuates proportionally to the distance due to resistance in the wire.

    Fibre has low noise, low attenuation and isn’t shared. Given our current state of physical knowledge it represents an ultimate technology.

    That makes it an oddity in economics, where generally we feel queasy about picking technological winners, but this really is a case where we can point to a technology and say ‘that one over there’, because we know what ‘perfect’ looks like. We should bite the bullet and just lay the cable.

  7. Actually with this in mind, it should be possible to do a back of the envelope model to answer your question. Say you have a neighbourhood with both fibre (with line rental) and wireless available. Give each user some utility function based on available bandwidth and cost. Come up with reasonable figures for available spectrum bandwidth from ACMA as well as signal to noise based on distance, plug that into shannon’s equation and that is the maximum datarate to be shared between all your wireless users.

    Set the fibre bandwidth to double every few years or so, a not unreasonable assumption, given that the limiting factor is a box at both ends, that can be changed by changing ISP. The model should go to equilibrium pretty quickly, and should give you some idea of whether there is a clear winner.

    I’m not an economist, but I’d be interested in the result.

  8. The last time I got into a heated discussion over the NBN, it turned out halfway through the thread that the bloke who was relentlessly arguing that wireless would soon overtake fibre also believed that the wireless broadband in his house was delivered from an antenna tower miles down the road.

    Caf’s “don’t really understand the mathematics” crowd isn’t hesitant about contributing to the NBN debate.

  9. Stephen: Certainly – that kind of “microcell” / “nanocell” set up is one way of increasing total available wireless bandwidth. It’s a form of spatial division – by using shorter-range transmitters you can pack more of them into a given area (if you have one powerful transmitter for an entire suburb, the available bandwidth has to be divvied up between more users than if you instead have 500 weak transmitters, one on each street corner). .

    As to whether it makes sense to replace the “last 25 metres” with low-power wireless, that’s a more easily analysed economic question. The answer might well differ between new estates and established suburbs, or depending on whether the existing house leadins are aerial or underground.

  10. Thanks for the useful comments. I started the post saying that Phil’s case was strong – your comments have convinced me that technically it is a lot stronger than I had thought!

%d bloggers like this: