May 30, 2003

Is the Radio Spectrum "Scarce"?

The Economist muses about whether the radio spectrum is "scarce" any more:


Economist.com: Economics focus

Freeing the airwaves
May 29th 2003
From The Economist print edition


Should radio spectrum be treated as property, or as a common resource?

WHAT is the best analogy for radio spectrum? Is it, as most people intuitively believe, a palpable resource like land, best allocated through property rights that can be bought and sold? Or is it, thanks to technological progress, more like the sea, so vast that it doesn't need to be parcelled out (at least for shipping traffic), in which case general rules on how boats should behave are enough to ensure that it is used efficiently.

Wireless folk have been discussing these questions for some time. Now, regulators are starting to take an interest, because increasing demands for wireless services require more efficient use of the spectrum. Earlier this month, America's Federal Communications Commission (FCC) decided to allow leasing and trading of frequency licences—the property model—as a first step towards establishing a market in radio spectrum. However, when regulators meet for the World Radiocommunication Conference in Geneva, starting on June 9th, they will try to harmonise their plans to expand the part of the spectrum that can be used without a licence, treating it as a common resource.

These two different regulatory models are already competing across the airwaves. On the one hand, telecoms companies have spent vast sums on licences for third-generation (3G) wireless services—but are facing serious financial and technical obstacles to building networks. On the other, there are already many wireless local access networks, called WiFi, which operate in unlicensed spectrum—and are growing at a phenomenal rate.

The question of how best to allocate spectrum is not new. Over 40 years ago Ronald Coase, who won the Nobel prize for economics in 1991, argued that there is no reason why spectrum should be treated differently from, for example, land. Both are scarce—so a market is the best way to allocate their use. Although this seems blindingly obvious today, it took the FCC more than two decades to start auctioning radio frequencies.

The debate has since moved on. Auctions alone are now considered unsatisfactory, because they do not change the traditional structure of spectrum allocation. And even after last month's reforms allowing leasing and trading, the FCC remains a dirigiste bureaucracy which decides, in most cases, how the spectrum is divvied up, who gets which slice, and for what use.

So what is the best way to replace this command-and-control regime? Proponents of the property approach want to create, as soon as possible, a market in which rights to spectrum blocks can be freely traded—rather in the way that pollution rights now are. Some have already drawn up plans for a “big bang”: a giant simultaneous auction of as much spectrum as possible.

Hold hard, say the advocates of common access. If spectrum were scarce by some law of nature, they argue, selling licences would certainly be the best solution. But in fact it is scarce only in terms of old, clunky technology. When radio equipment needed “channels”, defined by frequency and power, to allow communication without interference, airwaves were indeed a scarce resource.

Now, however, thanks to the dramatic decline in the cost of computer power, wireless devices are far cleverer, meaning that they can use spectrum more efficiently and are more tolerant of interference. They are able to communicate over a broad range of frequencies at once (this is called “spread spectrum”), to help each other out (“mesh networks”) and to adapt to the local environment (“agile radio”). Instead of creating a spectrum market, argues Yochai Benkler, a law professor at New York University, it should now be possible to rely on the market in smart radio equipment without anybody having to control the airwaves.

Technological progress is not the only reason why spectrum markets would be a second-best solution, Mr Benkler argues. For one, they are likely to come with high transaction costs. If spectrum is priced efficiently in an increasingly dynamic wireless world, the necessary overhead in network management and metering is likely to be quite costly. Innovation could suffer as well: rights holders could ignore technological improvement just because it does not fit their business model. With spectrum as commons, anybody can innovate, as users do on the internet.


Keeping the options open

Despite their differences, the two camps agree that they do not have enough hard data to bet everything on one regime: they must experiment with both. David Farber and Gerald Fulhaber, telecoms professors at the University of Pennsylvania, for instance, want a big-bang auction. But they also want to let others use spectrum freely, as long as they do not “meaningfully” interfere with the owner's right to a clear broadcast.

Despite its recent move toward a spectrum market, the FCC too prefers a hybrid approach, saying in a recent report that “no single regulatory model should be applied to all spectrum”. Early last year, it authorised systems using a technology called ultra-wideband to operate at very low power to avoid interference. The agency is also looking into expanding the part of the spectrum for which no licence is needed.

If experiences in other areas of technology are any guide, there is a good chance that both approaches will be around for some time, although the commons solution may eventually come to dominate. The internet, at least from the perspective of the end-user, is a common resource, with bandwidth allocated on a first-come-first-served basis. In software, the commons is growing, in the form of free open-source programs developed by volunteers.

Technology may thus help to create markets; but it also makes some of them obsolete. In this case it has turned land into sea, metaphorically speaking. To draw a historical parallel: the development of better ships did not lead to parcelling up the world's oceans but to something called free trade.

Posted by DeLong at May 30, 2003 10:37 AM | TrackBack

Comments

Where treating spectrum as a commons gets sticky are at least two points:

a) Each broadcaster raises the background noise floor for everyone else. This is a tiny, but real, cost on everyone else who wants to use the same area of spectrum. "Each action imposes a minor negative externality on everyone else" sounds an awful lot like pollution, and enough minor pollution can add up to something major.

b) This effect is particularly serious in certain key areas of spectrum, where massive amounts of sunk equipment cost mean adapting to a raised noise floor (even just leakage from users of different parts of the spectrum) is prohibitive. An obvious example is GPS, which already uses spread spectrum to get through with very low power levels. It imposes an implausibly huge burden to require the GPS constellation to adapt to more noise with less than 10-15 years of warning time before the adjustment in the constellation must be completed. Even if new satellites are designed, built and launched, significant increases in power require huge increases in satellite and launch costs. (Rememeber, 10dB of power is a factor of 10 more power).

Posted by: Tom on May 30, 2003 04:56 PM

Tom has nailed it. There is a finite amount of bandwidth available, regardless of how it is used; spread-spectrum technology alters the use, perhaps potentially improving allocation, but not the finitude.

Surely that most basic information theory result, Shannon's coding theorem, is known to economists? Prof. de Long?

Posted by: Randolph Fritz on May 31, 2003 12:34 AM

Tom has nailed it. There is a finite amount of bandwidth available, regardless of how it is used; spread-spectrum technology alters the use, perhaps potentially improving allocation, but not the finitude.

Surely that most basic information theory result, Shannon's coding theorem, is known to economists? Prof. DeLong?

Posted by: Randolph Fritz on May 31, 2003 12:35 AM

"Tom has nailed it. There is a finite amount of bandwidth available,..."

How many points are on a line? The electromagnetic spectrum is only limited by the technology used within it.

The more overlap into adjacent frequencies, the fewer channels can be fit into the spectrum. The less overlap, the more frequencies can be fit in.

If one has sufficient computational power to ignore all those transmissions that don't fit the "profile" one is looking for, I don't see how the electromagnetic spectrum is limited.

Posted by: Mark Bahner on May 31, 2003 04:07 PM

Mark is right. If current experiments in Ultra-wideband transmision pan out, it may no longer be tenable to talk about "limited spectrum". With UWB, your transimissions are indistinguishable from background noise to anyone not "tuned in" to your encoding scheme. Combine that with robust "ad hoc peer routing" architectures, and future wireless networks might be simply part of the air, pervasive and free to all...

Posted by: jimbo on June 1, 2003 09:26 AM

My understanding is that the truth is a mixture of the comments above.

>If one has sufficient computational power to ignore all those transmissions that don't fit the "profile" one is looking for, I don't see how the electromagnetic spectrum is
limited.

One limit of this is that particular approaches only allow so many "profiles." Another limit is my original point a) that if you are not using such a family of "profiles," then one approach will cause neglible amounts of interference for other approaches, but add enough "negligible amounts of interference" and the total interference can be quit substantial.


>Mark is right. If current experiments in Ultra-wideband transmision pan out,....

Mark & Jimbo are close to right, *for new equpiment*. A key issue is that deploying new equipment that takes advantage of all these sorts of new techniques can cause interference with old equipment. Essentially, by approving new equipment [that admittedly works much better for users who adopt], one can be implicitly mandating the obsolescence of old equipment. When that "old equipment" includes several billion dollars of satellites, plus zillions of user receivers, the "ouch" factor becomes quite noticeable.

Posted by: Tom on June 1, 2003 05:07 PM

"one approach will cause neglible amounts of interference for other approaches, but add enough "negligible amounts of interference" and the total interference can be quit substantial."

I guess that's what the experiments need to figure out: just how much does UWB "raise the noise floor"? I haven't been closely following it, but I'm not sure anyone knows for sure yet...

Posted by: jimbo on June 1, 2003 08:34 PM

UWB (if possible) would raise the noise floor by adding more "B" (bandwidth), of course.

"How many points are on a line?"

The key point is not "how many points?" but rather "how many can points can be distinguished?". And that number, as Claude Shannon (a student, I find, of Vannevar Bush) proved more than 50 years ago, is limited by noise, and noise is a property of any physical communications channel--there is an irreducible minimum based on molecular motion and (I think I recall) quantum effects.

It's a perpetual-motion machine argument. And I'm outta this one!

Posted by: Randolph Fritz on June 3, 2003 12:23 PM

"The key point is not "how many points?" but rather "how many can points can be distinguished?"."

Fair enough. Yours is a better characterization than mine.

"And that number, as Claude Shannon (a student, I find, of Vannevar Bush) proved more than 50 years ago, is limited by noise, and noise is a property of any physical communications channel--there is an irreducible minimum based on molecular motion and (I think I recall) quantum effects."

OK, again. But the key question is NOT whether there is some limit, but whether that limit is at all realistically approachable.

There is a limit to the number of water molecules available on earth, so that there is a limit to the amount of energy attained by fusion of hydrogen molecules from water. But this is NOT a limit that has any practical meaning, in terms of limiting the power humanity could get from fusion. We could power every single power-consuming device on earth for millions of years, and not even come close to running out of hydrogen from the water in a couple cubic kilometers of ocean.

With receivers that are computer-controlled to look for very specific frequency patterns generated by specific wide-spectrum sources, my understanding is that the amount of information that can be generated and received is far more than we presently transmit, and far more than we're likely to transmit in the forseeable future (I don't consider much beyond 30 years as "forseeable.")

"It's a perpetual-motion machine argument."

No, the I think the proper analogy is probably closer to the "fusion-limited-by-amount-of-water-on-earth" argument.

"And I'm outta this one!"

You haven't added much. What you've done is addressed the general issue without attaching ANY QUANTIFICATION. Without numbers, your argument doesn't hold much weight.

Your point that transmitters and receivers are limited by noise is completely valid. But the question is whether transmitter/receiver pairs that are matched to a wide range of frequencies (such as occurs in Ultra Wide Band transmissions) are ever limited by noise. My understanding is that they aren't:

http://www.ida.gov.sg/Website/IDAcontent.nsf/0/1856626048baf403c825698800267e26?OpenDocument

http://www.pulselink.net/ov_howitworks.html

http://www.timedomain.com/products/ourtech/faqs.html

Unfortunately, I also don't have any numbers, beyond: 1)UWB pulse lengths are incredibly short (1 picosecond), 2) their broadcast spectrum very wide, and 3) their power transmission is incredibly low (50 millionths of a watt, or less than 1/10,000th of the power of a cell phone)...so that a given section of the electromagnetic spectrum can be shared by an amazingly large number of UWB users.

I agree with Tom that interference with old, NON-UWB equipment is a problem. But it's my understanding that interference between UWB devices is essentially not an issue. It's my understanding that the electromagnetic spectrum is essentially unlimited, if used only by UWB devices.

Posted by: Mark Bahner on June 3, 2003 02:28 PM
Post a comment