How can the United States make better use of wireless spectrum? Broadly speaking, there are three categories of tools.
1. Engineering. Design and deploy transmitters and receivers that allow more data to be transmitted within given bands.
2. Allocation. Take spectrum away from owners who use it inefficiently and give it to more productive owners.
3. Pricing. Improve the incentives for users to make efficient use of spectrum
More discussion below the fold.
1. For over 15 years, I have heard claims that engineering can eliminate spectrum scarcity. At a recent conference that I attended, I heard one protagonist claim that the latest ideas in the lab could increase spectrum efficiency by a factor of 100 in five years.
At that same conference, another participant pointed out that a hundred-fold increase in efficiency in five years would be roughly the pace at which spectrum efficiency has increased over the past twenty years. Increases in efficiency only eliminate scarcity if they greatly exceed increases in demand, and that has not happened. What occurred to me is that claiming that innovative radio design will eliminate spectrum scarcity is like claiming that horizontal fracturing will eliminate energy scarcity.
The economic point is that the market is still going to clear at a non-zero price. Moreover, if prices stay at today’s levels, demand for spectrum (or for energy) will continue to rise, so that an increase in supply is required.
2. I am not happy with the FCC as a spectrum allocator. However, it is fair to say that they have done a lot over the past decade to try to make more private-sector spectrum available for general digital communication. At this point, a lot of the remaining wasted spectrum is controlled by government agencies, and the FCC is not authorized on its own to reallocate that spectrum. So better allocation will require a lot of political effort that might not yield a large gain. I am all for making that effort, but it will not be a panacea.
3. The demand for wireless communication varies by location and time of day. A wireless carrier is like an electric utility in that in order to keep its customers happy it wants to have enough capacity to meet peak demand. To me, this suggests that peak-load pricing could be a very powerful tool.
If I were a wireless carrier, I would offer my customers a plan that includes two “networks,” A and B. Physically, they are the same network, but virtually they are different. Network B is always available, but the per-megabyte charges for using network B are significant enough that you would think twice about using network B for more than just text messaging. Network A offers unlimited data and voice that is covered by the monthly fee, but in certain locations at peak times the user is going to be told that Network A is not available.
What I imagine happening in places like Manhattan is that as users start to notice that at certain times of day they often receive messages that network A is not available, they adjust their behavior accordingly. At those times of day, you start to see fewer movie downloads and photo uploads.
Questions about my dual-network idea:
Can it be made sufficiently consumer-friendly? For example, suppose I turn on my navigation system when network A is available, and halfway to my destination I hit a spectrum congestion point. As a consumer I don’t want to be told that all of a sudden I either have to switch to expensive network B or else turn off the navigation system.
Can it be made administratively low-cost? If operating the system uses more spectrum than it saves, then it does not help. And I suspect that the more consumer-friendly you want to make it, the more spectrum it will use.
Can the fears of “net neutrality” advocates be overcome? The term “net neutrality” means different things to different people. To some extremists, any price differentiation of any kind is too far down the slippery slope that leads to carriers telling you what content you can and cannot access. If they were logically consistent (which is asking a lot of that crowd), they would argue that the old “free nights and weekends” plans violate net neutrality, because they discriminate against people who need to communicate during weekdays.
In general, I think that “net neutrality” is a banner that attracts some nuts who have bad cases of what Bryan Caplan calls anti-market bias. They just think that markets are inherently unfair, and that “people” should not have to pay “corporations” for anything. That sounds uncharitable, but if you had heard such people at the conference you might think otherwise.
Anyway, I think that most reasonable people, including many advocates of net neutrality, would not object to peak-load pricing as a spectrum management tool.
My understanding of network neutrality is that all content is treated the same, not that quantity or time could not be treated differently.
One of the implications of this is that if customer has payed for a certain amount of data (at certain time periods even) then the carrier could not go out and charge certain sources of data to carry it to the end user (who has already payed for that data carriage). We see this in the big telecos charging netflix to carry their data, but not other data sources. If the end user has payed for that data carriage already, net neutrality would demand that the carrier could not discriminate what data the end customer actually uses it for.
“demand load” and “congestion load” pricing only help if the consumer has (a) the ability to control the time/route of their consumption and (b) the ability to do essential tasks while excercising this control.
(a) turns out that a great deal of spectrum consumption is beyond control – it’s automatic processes that transfer data in the background, incoming messages of all sorts that need to be delivered timely before the need for response can be decided, and so on. a pure “sender pays” scheme would change this, but in the US at least you are paying for bandwidth to receive various things.
(b) if the task at hand has to be done during business hours, or has to be done during evening entertainment hours, demand will be hard to shift. so while I can run the dishwasher and cloths dryer in the middle of the night with little hassle (and in fact do), that won’t work for watching the daily news or for sending a work report that somebody needs to reply to today.
Congestion pricing may matter somewhat at the margin, but it won’t dramatically affect the basic capacity need.
The biggest problem with your proposal is that it is very user-unfriendly. For most people, tech is very binary: it either works, or it doesn’t work. Your phone either connects, or it doesn’t connect. As well, people don’t pay attention to popups. They click ‘Yes’ without reading them. So a popup like “Network A is unavailable. Switch to Expensive Network B” will be clicked on with Yes automatically, and then charges will be disputed at end of month, and the media will run stories about the “evil phone companies tricking Grandma into paying thousands to see her baby granddaughter’s photos”.
Your proposal feels a lot like International Roaming charges. However, where roaming charges are predictable, your charges are not. And despite the predictability of roaming charges, you hear lots of stories of people accidentally running up huge bills when they go out of the country for a day or two.
Finally, from a technical point of view, I don’t understand what you mean by “Physically, they are the same network, but virtually they are different.” If they are the same network physically, a message on Network A adds to the congestion on Network B. You effectively only have one Network, where the price per megabyte changes from zero to X at different times of the day. I’m not really sure what logically separating the network into A and B gets you.
You could just as easily say:
There’s a network. Most of the time, sending data costs $X/mb (where X could be zero). However, when the network is congested, the price jumps to $Y/mb. The network protocol is capable of informing the device of the current state and price of the network, and the device may enable or disable connectivity based on that state and price.