Thursday, October 04, 2012

Three meanings of "spectrum efficiency"


“Efficiency” is a word of power, chanted when someone wants to bewitch an audience with the potency of economics. It’s often used in wireless policy, and I’ve realized that even when “spectrum efficiency” isn’t purely a fetish and is used to refer a ratio of input divided by output, the amounts compared depend on whether the speaker is an engineer or economist.

About a year ago I argued that spectrum efficiency metrics are meaningless. A recent TPRC paper with Ljiljana Simic, Petri Mähönen and Marina Petrova is more nuanced, and argues that spectrum efficiency, even in the narrow engineering sense, is a multi-dimensional problem and that fixating on one or two metrics is misleading (paper, slides). Claims for spectrum efficiency are usually little more than marketing – sometimes political, sometimes commercial, sometimes academic – masquerading as science.

Engineers understand and admit that spectrum efficiency, in its technical sense, is not a simple binary comparator that demonstrates unequivocally which of technology or spectrum use is more efficient than another, and therefore to be preferred. If they’re engaged in making policy choices they may try to shift the debate to more defensible ground, arguing that what really matters is “efficient use of spectrum”.

The terms “spectral efficiency” and “efficient use of spectrum” are often used interchangeably, on the basis that spectral efficiency could be taken to be a reasonably accurate estimate on how efficiently one is using spectrum. However, as the 2011 FCC TAC white paper on spectrum efficiency metrics pointed out, there are always “additional efficiency considerations” (its terminology) like device size, response time, quality of service, and build-out and operating costs that aren’t captured in spectral efficiency metrics.

There thus seem to be two broad senses in which “efficiency” is used: purely as abracadabra, and as a ratio. As an invocation, “efficiency” is used to conjure up and connote goodness. In this sense, “efficient use of spectrum” amounts to saying merely “the best use of spectrum”, where “the best” is short for “what I think is the best”. The magical power comes from invoking one of their favorite words of the gods of economics – just as the ancient gods feasted on the smell of a burning sacrificial offering, the gods of economics, and their devotees, smile when they hear the word “efficiency”.

Among technical types, the more common usage is efficiency as a ratio of output divided by input – at least, it’s what they purport to be talking about.  Some confusion arises between disciplines, though, because engineers and economists put different things in the numerator and denominator. For engineers, efficiency is usually data throughput divided by frequency bandwidth, or sometimes bandwidth multiplied by geographic area. For economists, efficiency is surplus divided by cost or investment.

In think one can see this tension in the FCC TAC white paper (written by engineers). The preferred metrics are all of the basic type “information bits per second divided by hertz of allocated spectrum (and sometimes geographic service area)”; the “additional efficiency considerations” include the economic considerations like reliability and security (output measures). and equipment cost and infrastructure investment (input measures).

To torture the religious analogy a bit more, one might think of the policy process as the engineers raising a bull called Efficiency, the economists slaughtering it and sprinkling it with their ideas, and the advocates and regulators burning it on the sacrificial pyre of politics…

One might be able to use textual analysis to put some flesh on the bones of this speculation, as I did with the "metaphors for spectrum" paper presented at the Dublin Dyspan. One might pick some canonical texts and then analyze how the term efficiency is used: ITU-R SM.1046-2, the CSMAC and TAC papers on efficiency, and some tech company CTO speeches for the engineering view; papers by Coase, Cramton, Klemperer, Hazlett and so on for auction theory and the economic perspective; and statements by FCC and EU Commissioners as well as rulemakings to represent the political discourse.

No comments: