Yochai Benkler delivers a talk that promises to present an analysis of the economic effects of commons and property models of spectrum allocation.
Right now, we have an incumbent regulatory system of command-and-control regulation. Both property and commons models are radical in comparison.
Spectrum is not “used up” — so you can’t optimize “spectrum use.” You’re optimizing communications. We’re optimizing the capacity of users to communicate information without wires.
There’s no reason for policy to aim at a false target (how we use spectrum best) imperfectly correlated to the actual target (how we communicate best).
The three paramaters for analysis:
- Equipment cost (total cost of the system)
- Displacement cost (how many communications, and of what value)
- Transaction/administrative cost
Equipment cost: in a commons model the machine capitalizes the value of free communications over the lifetime of the equipment. Most of the capital investment is done at the edges.
Property models make equipment optimized by pricing usage, but you get more investment in the network.
Displacement: in the old model, every communication displaced another communication. Destabilized by the declining cost of processors and the theoretical changes to Shannon’s Law and multiuser information channel.
Instead of adding infrastructure to gain capacity, you add users. The question is how to design equipment such that users add capacity — (Ed: It’s the sheep that shits grass!)
Processing gain and cooperation gain mean that there is no fixed “amount of spectrum” necessary for communication. You can’t define the displacementeffect independent of the configuration of the actual devices invilved, theiir geography and the actual devices potentially displaced.
In a commons model, there is value in investing in very good receivers, but in a property network, there is no such incentive, since you can command users of your property — your spectrum — to silence themselves while you speak and listen.
Will the smaller number of communications cleared by a property system be worth the added reliability of those comms?
Transaction costs: You’ve got to figure out all the variable in every communication (i.e., where to trasnmit it, on what frequency, etc) — in a property model, you’ve got to send out bids to all the possible paths and calculate the optimal price/quality tradeoff. A single communication will require dozens of transactions in multiple places with multiple people. Given how dynamic the spot market for spectrum will be, the tranaction cost will be too high to be practical. To the extent that property models use statistics to generate average pricing, you will sacrifice any of the efficiencies you gain from actual markets.
This is a market in infrastructure versus market in equipment, not market versus nonmarket.
The long-term solution is mostly commons with some property for Qualilty of Service mechanisms. We’re not far along enough to determine now what we need — we should try both for ten years and then revise it all.