Interesting People mailing list archives
getting to be interesting -- a scarcity-creating model
From: David Farber <dave () farber net>
Date: Sun, 15 Jun 2008 14:41:01 -0700
________________________________________ From: Bob Frankston [bob37-2 () bobf frankston com] Sent: Sunday, June 15, 2008 5:26 PM To: David Farber; 'ip' Cc: Gerry Faulhaber; dpreed () reed com Subject: RE: [IP] a scarcity-creating model Congestion pricing leads to a reduction in capacity when there’s a danger of abundance. We need to be careful about the Malthusian/finite pie model – makes it easy to do the math but is a false premise. The costs are not directly related to capacity because we don't have a fixed size pie we're divvying up. The capacity settings are not fundamental -- they are the result of policy and not physics. It’s not just physics – but also what I’ve been calling the opportunity dynamic – finding value in what we have rather than being dependent upon a single solution or, worse, silo dependencies. (http://www.frankston.com/?name=InternetDynamic). There are many ways to increase effective capacity. The scarcity assumption is the same kind of fallacy as assuming that we are limited by the laws of thermodynamics when we don't have closed system or in confusing Shannon's mathematical model with the physics (and innovation) in a particular situation. It’s also like assuming demand and supply are in competition rather than recognizing demand creates supply and supply creates demand when you’re not in a closed system – at least to a first approximation. This is why congestions pricing leads to a reduction in capacity -- in the carrier's own words -- http://www.frankston.com/?name=AssuringScarcity they must guard against abundance lest the prices collapse. This is precisely what would/should happen if capacity could be increased without bounds (and, again, that includes finding more value in the capacity we already have). Furthermore users can be very flexible and adaptive – they automatically adapt to what is available without a direct link to the millisecond pricing. Given that no player has sufficient control to deliver on a promise in return for more money we’re left with just-so stories and not real price incentives. Remember also that local ISPs are just customers and don’t operate any more than a smidgen of the network. The extreme case being the home network which tends not to be capacity constrained since gigabit paths are very inexpensive. If there is a local constriction that’s an anomaly and mustn’t be the basis for global policy. Attempt to reduce demand is a strange goal – especially when it is totally unnecessary and is far more efficient than spending untold resources working around artificial constraints. http://www.frankston.com/public/rss.xml -----Original Message----- From: David Farber [mailto:dave () farber net] Sent: Sunday, June 15, 2008 16:37 To: ip Subject: [IP] a pricing model?? ________________________________________ From: Gerry Faulhaber [gerry-faulhaber () mchsi com] Sent: Sunday, June 15, 2008 4:03 PM To: David Farber Cc: brett () lariat net Subject: Re: [IP] ALSO MUST READ NYTimes.com: Charging by the Byte to Curb Internet Traffic "People seem to be missing the point." [for IP, if you wish] Whew! Somebody finally got it right! When all the costs are capacity costs (as in the Internet), then the economically most efficient pricing (in theory) is what Bohn calls "spot" pricing, or what economists call "Vickrey" pricing (after Nobel laureate Willliam Vickrey, Columbia Univ.), sometimes called "congestion" pricing (see Wikipedia): the price for use of a facility is zero (assuming running costs are zero) if demand is less than capacity, but increase during times of peak use when demand is great enough to use all the capacity available. All customers who use the facility during periods of congestion pay this congestion price (BitTorrent users and people that want to browse the Web alike), and zero otherwise. The congestion price should be sufficient to reduce demand so that demand just equals capacity during this period, so that all customers willing to pay for the congestion they cause are able to use the facility. Customers not willing to pay can postpone their use of the facility. If congestion pricing leads to revenues greater than the cost of capacity, then the owner of the facility has an incentive to invest in more capacity, and will do so up to the point where marginal revenues are just equal to the marginal cost of capacity. In theory, the congestion price is only imposed when capacity is congested. While this is theoretically correct, it has proven to be difficult (if not impossible) to implement in practice. Since we are never sure exactly when congestion will occur, there must be some way to inform customers what the price is at this moment, and they must be able to respond to this information in real time. And what happens when someone opens a BitTorrent download when demand is slack (zero price) but the price increases twenty minutes into a two-hour download, maybe to $1/Mb? Approximations have proven useful, such as peak load pricing. In this model, the owner sets prices by time of day, under the assumption that peak demand always occurs at the same time (this concept goes back to Steiner in the 1950s), which is likely false. Further, in order for time of day pricing to correctly ration customers' usage, they must all be aware of when the prices change and by how much they change. Price-elastic customers were pretty good at this in the old telephone world (where time of day pricing was standard), but not everyone. Cell phone firms have opted for a "bucket of minutes" approach, which is a form of usage-based pricing that customers seem to accept. It only controls peak demand if one makes the heroic assumption that across all customers, total usage and peak usage are highly correlated. Certainly not true in general, but I suspect that as an approximation it's not too bad. I also suspect that this will be the direction that ISPs will take: do something that customers have shown a willingness to accept (buckets of minutes), even if it is only a fair approximation to the theoretically correct pricing. If I were a broadband ISP (Brett Glass, you can chime in here), this is what I would do. I would transition by giving customers who use less than (say) 4 Gb a month a price reduction (for the lowest bucket), which could well cover 80% of customers, and the rest would get a healthy price increase. Theoretically perfect? Far from it. But a reasonable first move away from the unsustainable All You Can Eat pricing we have now. Prof. Gerald Faulhaber (Emeritus) Wharton School and Law School University of Pennsylvania ----- Original Message ----- From: "David Farber" <dave () farber net> To: "ip" <ip () v2 listbox com> Sent: Sunday, June 15, 2008 1:08 PM Subject: [IP] ALSO MUST READ NYTimes.com: Charging by the Byte to Curb Internet Traffic "People seem to be missing the point." ________________________________________ From: Roger Bohn [Rbohn () ucsd edu] Sent: Sunday, June 15, 2008 11:13 AM To: David Farber Cc: Bob.Frankston () indigo pobox com; "[bob37-2 () bobf frankston com]"@indigo.pobox.com; Michael.O'Dell () indigo pobox com; "[mo () ccr org]"@indigo.pobox.com Subject: Re: [IP] MUST READ NYTimes.com: Charging by the Byte to Curb Internet Traffic "People seem to be missing the point." Regarding the reaction to the NY Times article and the whole subject of charging by the byte, let me point out that the correct way to charge for scarcity of bandwidth is ALSO something that many readers of this list will distrust.
From: Michael O'Dell [mo () ccr org]
Since bit *rate*, not bit mass, is the instantaneously-exhaustable
resource in packet network, if they were actually worried about
network engineering, they'd be going to the burstable charging
model which is known to worth both technically and economically.
it relates the charges directly to the exhaustion of the
finite resource - bit *rate*, not bit mass. Mr. O'Dell is correct about rate, but the scarce resource here is _congestion_ flows, which are highly variable. The correct (economically, and in my modestly informed opinion technically) way to charge for these is with some form of spot pricing, i.e. prices that change in real time and with location. When there's no congestion, no matter how much bandwidth someone currently uses the charge should be zero. Conversely, when your neighbors are running real-time movies via IP, both you and they should be charged congestion fees for whatever each of you is doing. Even if you did not "cause" the congestion, your usage is exacerbating it for everyone. The easiest analogy is to cellular phones, which have gone as far as a two-level time-of-day price (free/ not free), but otherwise stayed away from a "burstable charging model." Electricity sellers are starting to play with spot prices. But in general they are viewed as out of the question for consumers because of the uncertainty and variability they introduce. Your monthly payment becomes even harder to predict than with a pure charge for bytes.* So like Network Neutrality, this is a case of "be careful what you ask for...." How many of IP's readers would like to give some kind of real-time pricing power to their ISP? Imagine the problems of auditing your bill to ensure that it was correct, for example. Solvable technically, but when there is a rapacious quasi-monopolist writing the bills it would require a lot of trust. So a time-of-day based surcharge for bytes seems like a reasonable compromise between tractability and theoretical optimality. This is not what Comcast is proposing, though, which lends credence to the theories that they have a very different agenda than what they claim. Long-term solution: We need a third source of high-bandwidth to the home - probably nothing less will get the ISPs to behave. This would also, most likely, solve the net neutrality problem without a lot of dangerous micro-regulation of what behavior is acceptable. Roger *(Doing spot pricing economically correctly for the Internet would be harder than for cellular, because of end-to-end issues, which can lead to a lot of nonsense about capacity reservations, "fair queuing," and the rest. Hans-Werner Braun, KC Claffy, and I wrote a paper about end-to-end spot pricing in the mid 90s. If they are only worried about local congestion, though, as the Comcasts of the world imply in their PR, then backbone and other-end congestion can be ignored.) ------------------------------------------- Archives: http://www.listbox.com/member/archive/247/=now RSS Feed: http://www.listbox.com/member/archive/rss/247/ Powered by Listbox: http://www.listbox.com ------------------------------------------- Archives: http://www.listbox.com/member/archive/247/=now RSS Feed: http://www.listbox.com/member/archive/rss/247/ Powered by Listbox: http://www.listbox.com ------------------------------------------- Archives: http://www.listbox.com/member/archive/247/=now RSS Feed: http://www.listbox.com/member/archive/rss/247/ Powered by Listbox: http://www.listbox.com
Current thread:
- getting to be interesting -- a scarcity-creating model David Farber (Jun 15)